Mapping the Landscape of Intervention Programmes to Mitigate Mis- and Disinformation

1 2 Please cite this report as: Azadi, T., d’Haenens, L., De Nolf, A., Ponte, C., Luna, E., Tomczyk, Ł., Donoso, V., Torres da Silva, M., Batista, S., & Żegleń, M. (2025). (2025). Mapping the Landscape of Intervention Programmes to Mitigate Misinformation and Disinformation: A Scoping Review. PRODIGI, KU Leuven. Disclaimer PRODIGI is funded by the European Union, under Grant Agreement no. 101182849. Views and opinions expressed are, however, those of the author(s) only and do not necessarily reflect those of the European Union. The European Union cannot be held responsible for them. 3 Mapping the Landscape of Intervention Programmes to Mitigate Misand Disinformation Work package 1 – Deliverable 1.1 Submission date: 19 December 2025 Lead beneficiary: KU Leuven Authors: Tania Azadi, Leen d’Haenens, Ans De Nolf, Cristina Ponte, Estrella Luna, Łukasz Tomczyk, Verónica Donoso, Marisa Torres da Silva, Susana Batista, Magdalena Żegleń 4 Table of contents 1. Executive Summary ......................................................................................................................... 7 Scope and study characteristics .......................................................................................................... 7 Intervention types and inputs ............................................................................................................. 7 Activities and delivery ......................................................................................................................... 7 Outputs, outcomes, and effectiveness ............................................................................................... 8 Challenges and facilitators of implementation ................................................................................... 8 1. About PRODIGI .............................................................................................................................. 10 The PRODIGI project ......................................................................................................................... 10 Objectives...................................................................................................................................... 10 PRODIGI Consortium ..................................................................................................................... 11 2. Introduction .................................................................................................................................. 13 This report ......................................................................................................................................... 13 3. Methods ........................................................................................................................................ 14 Silvi.AI ................................................................................................................................................ 14 Eligibility criteria (Population–Concept–Context) ............................................................................ 15 Information bibliographic databases ................................................................................................ 15 Search strategy and execution .......................................................................................................... 15 Search strings .................................................................................................................................... 16 Operational steps .............................................................................................................................. 16 Study selection process (screening) .................................................................................................. 16 Data charting (extraction) ................................................................................................................. 18 Synthesis and presentation ............................................................................................................... 20 4. Results ........................................................................................................................................... 20 5.1 Characteristics of the included studies ....................................................................................... 20 Study setting ................................................................................................................................. 20 Study design .................................................................................................................................. 21 Presence of comparison between groups .................................................................................... 21 Data collection methods ............................................................................................................... 22 Sample size characteristics ........................................................................................................... 23 5 5.2. Intervention types, scope, and aims: distinguishing AI literacy from other literacy interventions .......................................................................................................................................................... 25 AI literacy interventions ................................................................................................................ 25 Other literacy interventions (media, information, and misinformation literacy) ........................ 25 Comparative perspective .............................................................................................................. 26 5.3 Human resources, technological inputs, and materials characterising the interventions .. 27 Human resources .......................................................................................................................... 27 Technological resources ................................................................................................................ 27 Materials and pedagogical resources ........................................................................................... 27 5.4. Intervention activities and delivery characteristics ................................................................... 29 Activity types and descriptions ..................................................................................................... 29 Delivery modes and group formats .............................................................................................. 29 Session count, duration, and intervention period ........................................................................ 29 Facilitation models ........................................................................................................................ 30 Theoretical frameworks ................................................................................................................ 30 Frameworks underpinning AI literacy interventions .................................................................... 30 Comparative observations ............................................................................................................ 31 Gaps and limitations in theoretical articulation ........................................................................... 31 5.5 Intervention outputs and reach .................................................................................................. 31 Number of participants reached ................................................................................................... 32 Number of sessions delivered ....................................................................................................... 32 Products and materials developed ............................................................................................... 32 5.6 Intervention outcomes ............................................................................................................... 33 Measured outcomes ..................................................................................................................... 33 Self-reported outcomes ................................................................................................................ 34 Comparative patterns and methodological gaps .......................................................................... 34 Outcomes of digital and media literacy interventions ................................................................. 35 Outcomes of AI literacy interventions .......................................................................................... 36 Comparative patterns and outcome gaps .................................................................................... 36 5.7 Outcome effectiveness: direction, magnitude, and strength of evidence ................................. 37 Direction of effects ....................................................................................................................... 38 Effect size reporting ...................................................................................................................... 38 Statistical significance ................................................................................................................... 38 Quantitative and qualitative evidence .......................................................................................... 38 5.8 Challenges and facilitators of implementing misand disinformation/AI literacy interventions .......................................................................................................................................................... 39 6 Facilitators of implementation ..................................................................................................... 39 Challenges to implementation ...................................................................................................... 40 5. Discussion...................................................................................................................................... 41 6. Conclusion ..................................................................................................................................... 42 Actionable recommendations for the PRODIGI interventions ......................................................... 43 1. Integrate AI literacy and digital/misinformation literacy by design ......................................... 44 2. Prioritise facilitated, active learning formats ............................................................................ 44 3. Design for depth over single-exposure effects ......................................................................... 44 4. Align outcomes with intervention aims and use mixed measures ........................................... 44 5. Embed vulnerability and context explicitly ............................................................................... 45 6. Support facilitators through targeted capacity building ........................................................... 45 7. Design for reuse, adaptation, and sustainability ...................................................................... 45 8. Combine quantitative and qualitative evidence in evaluation ................................................. 45 7. References .................................................................................................................................... 46 8. Appendices .................................................................................................................................... 48 A. Code Book, V.2 .............................................................................................................................. 48 Group 1: Study_Characteristics .................................................................................................... 48 Group 2: Intervention_LogicModel .............................................................................................. 48 Group 3: Inputs (Logic Model – what goes in): ............................................................................. 48 Group 4: Activities (Logic Model – what is done): ........................................................................ 48 Group 5: Outputs (Logic Model – immediate products): .............................................................. 49 Group 6: Outcomes (Logic Model – changes observed/measured): ............................................ 49 Group 7: Outcomes_Effectiveness ................................................................................................ 49 B. Overview of included studies ........................................................................................................ 51 7 1. Executive Summary This scoping review synthesises evidence from 132 studies on interventions to address misinformation, disinformation, and AI literacy. The review maps the characteristics, design features, implementation conditions, outputs, outcomes, effectiveness, and challenges of these interventions, with particular attention to distinctions between AI literacy and broader digital and media literacy approaches. Scope and study characteristics Most of the available evidence comes from high-income Western countries, especially the United States and Western Europe. Interventions are usually delivered in online settings or within formal education systems, likely because these environments are easier to control and scale. Research in this area mainly relies on randomised and quasi-experimental designs, with a focus on internal validity and measuring short-term outcomes. In contrast, studies employing longitudinal, qualitative, or implementation-focused approaches remain relatively uncommon. Study participants were most commonly students and adults recruited online. Far less research has focused on children, older adults, professionals, or people facing structural disadvantages. Although around one-third of studies included populations described as vulnerable, vulnerability was often assumed rather than clearly defined or examined in depth. Intervention types and inputs Interventions fall into two analytically distinct but partially overlapping categories: • Digital, media, and misinformation literacy interventions, which focus on content-level evaluation, critical thinking, and resistance to misinformation. • AI literacy interventions, which emphasise system-level understanding of algorithms, datadriven decision-making, and the ethical and societal implications of AI. Most interventions required a moderate level of resources. They typically relied on facilitation by educators or researchers, used widely available digital technologies, and employed teaching-oriented materials. Few interventions required advanced technical expertise or fully functional AI systems, highlighting a strong focus on feasibility and scalability. Activities and delivery Most interventions combined teaching with active learning approaches, such as guided analysis, scenario-based activities, and reflection. They were usually delivered in group settings with a facilitator, although a substantial minority used self-directed online formats. Overall, interventions tended to be low to moderate in intensity, most often consisting of a single session or a small number of sessions. Longer, curriculum-integrated programs were relatively uncommon. The extent to which interventions were grounded in theory varied. Digital literacy interventions were most often informed by media and information literacy frameworks, critical thinking theories, and inoculation theory. 8 In contrast, AI literacy interventions more frequently drew on socio-technical, ethical, and critical perspectives on technology. However, many studies did not clearly state the theoretical frameworks guiding their design. Outputs, outcomes, and effectiveness Reported outputs focused primarily on participant reach and the educational materials produced through the interventions, with most studies involving small to medium-sized samples. Although many projects generated reusable digital resources or tools, there was limited systematic reporting on their longer-term uptake or institutional adoption. Outcomes were most often assessed using short-term quantitative measures, relying heavily on self-reported indicators of attitudes and perceived competence. In contrast, more direct measures were used for knowledge and evaluative skills. Across both intervention types, the most consistent and robust effects were observed in improvements in knowledge acquisition and critical evaluation skills. AI literacy interventions were particularly effective in strengthening understanding of how AI systems work and increasing awareness of algorithmic bias and ethical issues, though their effects on trust in or attitudes toward AI were mixed and often reflected increased critical scepticism rather than acceptance. Overall, most studies reported positive effects on at least one outcome, typically small to moderate in size. However, effect sizes were inconsistently reported, and evidence of long-term change, behavioural impact, or real-world application remains limited. Where qualitative findings were included, they added depth by revealing reflective shifts and ambivalence that quantitative measures did not fully capture. Challenges and facilitators of implementation Implementation often worked well when supported by easy-to-use technologies, clear guidance from teachers or facilitators, hands-on learning, and existing teaching models. However, several difficulties were also reported, such as limited time, the complexity of some ideas, especially around AI, uncertain links to theory, and very ambitious goals compared to the short-term results that were measured. In interventions addressing misinformation, an unexpected effect emerged: while participants became more cautious about misinformation, some also became less confident in accurate, trustworthy news. Overall conclusions Taken together, the evidence points to a field that is methodologically sophisticated but gives relatively little attention to implementation. There is strong short-term evidence for gains in knowledge and skills, but much less insight into how durable these gains are, whether they transfer to other contexts, or how they translate into real-world impact. AI literacy and digital and media literacy interventions address different but complementary aspects of contemporary information challenges, pointing to considerable potential for more integrated approaches that combine content-focused critical skills with broader awareness of AI systems. 9 As such, future research would benefit from: • stronger theoretical articulation, • more diverse populations and contexts, • longitudinal and mixed-methods designs, • and greater attention to scalability, sustainability, and implementation conditions. From a scoping perspective, these findings suggest a field with encouraging evidence of effectiveness, but one that is uneven in the design and reporting of studies. There is clear room for improvement in enhancing comparability of results, assessing outcomes over longer periods, and combining quantitative and qualitative methods to provide a fuller understanding of what works and why. 16 Search strings Our search strings include the following: Concept 1: Misand disinformation: ("misinformation" OR "disinformation" OR "fake news" OR "false information" OR "information disorder" OR "rumor*" OR "hoax*" OR "conspiracy theor*" OR "deceptive content" OR "manipulated information" OR "fabricated news" OR "propaganda" OR "misleading information") Concept 2: Media and digital and artificial: ("media" OR "digital" OR "online platform*" OR "social media" OR "internet" OR “platform” OR "digital platform*" OR "news media" OR "mass media" OR "information technolog*" OR "technolog*" OR "AI" OR "artificial intelligence" OR "algorithm*" OR "automated system*" OR “chat gpt” OR “copilot” OR “open ai”) Concept 3: Literacy and skills: ("skill*" OR "literac*" OR "knowledge" OR "competenc*" OR "awareness" OR "capability*" OR "critical thinking" OR "digital literacy" OR "media literacy" OR "AI literacy" OR "information literacy" OR “platform literacy” OR “social media literacy” OR "analytical skill*" OR "evaluation skill*") Concept 4: Intervention and programs: ("intervention" OR "program" OR "initiative" OR "educational program" OR "training program" OR "curricul*" OR "campaign" OR "project" OR "workshop" OR "learning module" OR "awareness campaign" OR "school program" OR "community outreach" OR "capacity-building") Concept 5: Methodology: (“experiment*” OR “RCT” OR “randomised control* trial” OR “case control” OR “control group” OR “quantitative” OR “evaluat*”) Operational steps Operationally, the search strategy was translated and implemented across databases by adapting quotation marks, field tags (e.g., title/abstract/keywords), truncation, proximity operators, and subject headings where available to fit each platform’s syntax. A small pilot search was then conducted to confirm that the strategy retrieved known relevant publications, and the balance between sensitivity and specificity was refined iteratively in line with scoping review guidance. The final searches were restricted to English-language publications from 2021 to 2025 in accordance with the protocol, and complete documentation was maintained by saving the full search strings, the dates each search was run, and the number of records retrieved per database to ensure reproducibility and PRISMA-ScR–aligned reporting (Tricco et al, 2018). Study selection process (screening) The study selection process followed a transparent and reproducible workflow consistent with PRISMA-ScR (Tricco et al., 2018). Figure 1 presents the PRISMA-ScR flow diagram summarising the identification, screening, eligibility assessment, and inclusion of studies in this scoping review. The database searches yielded 4,036 records. After removing duplicates (n = 2,033), 2,003 unique records remained for title and abstract screening. Using the predefined inclusion and exclusion criteria, 1,800 records were excluded at this stage, and 202 articles proceeded to full-text screening. 17 During full-text assessment, the same eligibility criteria were applied, and a further 71 articles were excluded, with reasons documented. In total, 132 studies met the eligibility criteria and were included in the final analysis and synthesis. The study selection process is summarised in this PRISMA flow diagram (Figure 1). Figure 1: PRISMA-ScR flow diagram of the study selection process, showing records identified, duplicates removed, screening, full-text assessment, exclusions, and final included studies (N = 132) 18 All identified citations were first exported to an Excel file and then uploaded to the Silvi platform, where duplicate records were removed using automated checks. Ten reviewers were grouped into three screening teams (two groups of three and one group of four). Each team independently reviewed titles and abstracts against predefined eligibility criteria. Any disagreements were discussed within these teams, and when consensus could not be reached, two additional reviewers acted as independent adjudicators. The same independent screening process was used at the full-text stage, applying the same inclusion and exclusion criteria and systematically recording reasons for exclusion to ensure transparency. The entire selection process, from initial identification and de-duplication through screening and final inclusion, was documented using a PRISMA flow diagram adapted for a scoping review. Inter-rater reliability measures Inter-rater reliability (IRR) was assessed at both screening stages (title/abstract and full text) to quantify agreement between reviewers, identify sources of inconsistency in applying the eligibility criteria, and inform targeted calibration of the screening process (Table 1). Three complementary indices were computed: raw agreement (the proportion of identical decisions), pairwise PABAK (a prevalenceand bias-adjusted agreement coefficient), and Krippendorff’s alpha (a chance-corrected reliability coefficient suitable for multiple raters). Table 1: Inter-rater reliability (IRR) statistics for title/abstract and full-text screening phases Screening phase Raw agreement Pairwise PABAK Krippendorff’s alpha Title/abstract screening 0.8811 0.7990 0.43036 Full-text screening 0.5518 0.0475 0.1193 At the title/abstract stage, raw agreement was high (0.8811), indicating that reviewers reached the same decision for the large majority of records. The pairwise PABAK was also high (0.7990), indicating substantial agreement after controlling for decision prevalence and rater systematic bias. However, Krippendorff’s alpha was notably lower (0.43036), indicating only fair-to-moderate reliability beyond chance. Taken together, this pattern is consistent with a screening context in which “exclude” decisions were much more frequent than “include/uncertain” decisions (a common scenario in scoping reviews): raw agreement can appear high under such prevalence imbalance, while chancecorrected coefficients such as alpha may remain modest. At the full-text stage, raw agreement across reviewers was 0.5518, indicating moderate concordance in initial eligibility judgments. Chance-corrected agreement coefficients were low (pairwise PABAK = 0.0475; Krippendorff’s α = 0.1193). Full-text screening required detailed assessment of multiple eligibility criteria and often involved borderline cases, incomplete reporting, and interpretive judgments. Given this level of task complexity, some degree of disagreement between reviewers was expected and is a common feature of full-text screening in systematic reviews. Because all records were triple reviewed and discrepancies were resolved through consensus, these IRR estimates primarily describe variability in initial judgments rather than the reliability of the final inclusion decisions. The consensus-based approach functioned as an adjudication mechanism to ensure consistent application of eligibility criteria at the final decision stage. Data charting (extraction) For data charting and synthesis, we used the Silvi platform’s tagging functionality to support structured extraction from included full texts, following the scoping review principle that data charting 19 is an iterative process that can be refined as familiarity with the evidence base increases (Arksey & O’Malley, 2005; Levac et al., 2010; Peters et al., 2020a). We developed a codebook based on the Logic Model (W.K. Kellogg Foundation, 2004; McLaughlin & Jordan, 1999) to describe interventions and their intended pathways, and we operationalised each code with definitions and decision rules to support consistent application across reviewers. The codebook was developed through a combination of deductive coding (informed by the logic model and review questions) and inductive refinement (informed by insights gained during title/abstract and full-text screening), consistent with guidance that scoping reviews may iteratively adjust charting categories to better capture study characteristics and outcomes (Levac et al., 2010; Peters et al., 2020b). Two reviewers piloted the initial version on a sample of included papers (n = 5) to calibrate interpretation and further refine fields for clarity and completeness before it was adopted by the whole team and implemented in Silvi. After the codebook was integrated into the platform, Silvi’s AI-assisted feature pre-populated tags by scanning full texts and extracting candidate data elements. Every AI-generated extraction was reviewed by a human reviewer who approved, edited, or deleted entries to ensure accuracy and consistency. This resulted in a combined AI–human workflow in which automation supported efficiency while final charted data remained reviewer-verified. Ten reviewers contributed to data charting by extracting data directly and/or validating the AI-assisted outputs. The final codebook (Appendix A) captured bibliographic information (author, year, title, publication type, country/region), population and vulnerability characteristics (participant demographics, vulnerability definition/criteria, setting), detailed intervention characteristics (focus on media literacy/digital skills/AI literacy; misinformation/disinformation components; AI-related elements such as algorithms or chatbots; delivery mode; duration/intensity; implementers; tools/materials; setting and level of delivery; and comparators where applicable), study design and methods (qualitative/quantitative/mixed methods; evaluation design including RCT or quasi-experimental approaches when present; measures/instruments; and follow-up), outcome findings (outcome domains, direction of effects, effect sizes and statistical significance when reported, and qualitative themes where relevant), and implementation/context notes (feasibility, barriers/facilitators, equity considerations, and relevance to Belgium, Portugal, Poland, and broader EU transferability). An overview of all included studies can be found in Appendix B. 20 Synthesis and presentation The synthesis and presentation of results focused on mapping and describing the available evidence using a descriptive summary (e.g., counts by year, country/region, population group, intervention type, setting, study design, and outcome domains) and a structured narrative synthesis aligned with scoping review guidance (Arksey & O’Malley, 2005; Peters et al., 2020a) rather than undertaking metaanalysis, except in the unlikely event that a clearly homogeneous subset of studies justified quantitative pooling. We also organised results into evidence maps and cross-tabulations to show how intervention components related to target groups, settings, and outcome domains, which is consistent with recommendations that scoping reviews present results through both tabular mapping and narrative explanation to identify patterns and gaps (Peters et al., 2020a; Peters et al., 2020b). 4. Results 5.1 Characteristics of the included studies Across the 132 studies included in this scoping review, substantial variation was observed in geographical location, study setting, methodological design, and target populations. With respect to the country of implementation, the evidence base was strongly skewed toward high-income Western countries. A clear majority of studies (well over half of the total corpus) were conducted in the United States and Western Europe, including the United Kingdom, the Netherlands, and Germany. A smaller but non-negligible proportion of studies took place in lowand middle-income countries, including contexts in South Asia and Sub-Saharan Africa. However, these settings remained underrepresented relative to their population size and exposure to the issues targeted by the interventions. Study setting Interventions were implemented in a range of settings, with a pronounced concentration in digitally mediated environments. Approximately three-quarters of the 132 studies were conducted online, including web-based experimental platforms, social media environments, and purpose-built digital interfaces. These interventions typically exposed participants to stimuli such as messages, videos, interactive modules, or simulated content under controlled experimental conditions. The predominance of online settings reflects both the digital nature of the phenomena under investigation and the methodological advantages of online recruitment and randomisation. Educational settings constituted the second most prevalent intervention context, accounting for approximately one quarter of the studies. These interventions were primarily implemented in secondary schools and higher education institutions and were often embedded in classroom activities, curricular modules, or structured training sessions. School-based studies frequently relied on quasi-experimental designs that used intact classes or cohorts and were more likely to involve younger participants, including adolescents. 21 A smaller subset of studies, less than one tenth, was conducted in community-based or institutional settings outside formal education. These included interventions delivered through community centres, non-governmental organisations, libraries, or public workshops. Such settings were more commonly associated with interventions targeting specific population groups, including marginalised or underserved communities, and often emphasised applied or participatory approaches. In addition, a limited number of studies employed hybrid settings that combined online components with offline delivery. These interventions typically involve in-person instruction or discussion supplemented by digital materials, online exercises, or follow-up assessments. Hybrid designs were particularly evident in educational and community-based interventions seeking to balance scalability with contextual engagement. Finally, a small minority of studies did not clearly specify a discrete physical or digital setting, instead focusing on exposure to intervention materials without explicit contextual embedding. These studies were typically laboratory-style experiments or secondary analyses in which the setting was treated as analytically neutral. Overall, the dominance of online and educational settings underscores the field’s strong orientation toward controlled, scalable intervention environments. At the same time, the relatively limited representation of community-based and hybrid contexts highlights an opportunity for future research to examine intervention effectiveness in more ecologically embedded and diverse real-world settings. Study design The body of evidence was primarily based on experimental studies, reflecting a strong focus on testing interventions. Most of the 132 studies used randomised controlled designs, typically conducted online or in laboratory settings, in which participants were randomly assigned to one or more intervention conditions and compared with a control or standard information condition. A smaller but still meaningful group of studies used quasi-experimental designs that compared groups without complete randomisation, often because the research took place in real-world settings such as schools or community organisations, where intact classes or institutions served as comparison groups. Fewer studies relied on single-group before-and-after designs without a control group, mainly in pilot or early-stage projects aimed at exploring whether an intervention showed promise. Only a small number of studies used non-experimental approaches, such as cross-sectional or retrospective analyses, typically to examine patterns of exposure or implementation rather than effectiveness. Overall, most studies included some form of between-group comparison, often testing multiple versions of an intervention against a control group. While this emphasis on experimental and comparative designs supports firm conclusions about cause and effect, the limited use of long-term, mixed-methods, or implementation-focused studies means that less is known about how durable and transferable to other contexts these effects are, how interventions work in everyday settings, and how they can be sustained in practice. Presence of comparison between groups Most studies in our scoping review used comparative designs to evaluate interventions. In the vast majority of studies (approximately four out of five), interventions were assessed by comparing one group that received the intervention with another that did not, such as a control or standard information group. 22 These comparisons most often involved assigning participants to different groups and examining differences in outcomes between them, providing a straightforward way to assess whether the intervention had an effect. In some cases, studies tested more than one version of an intervention, allowing researchers to compare different messages, formats, or delivery approaches. By contrast, around one in five studies (approximately 20%) did not include a comparison group. These studies typically measured outcomes before and after participants were exposed to an intervention, or assessed outcomes only after the intervention. Such approaches were most often used in pilot studies or early-stage research, where the aim was to explore whether an intervention was feasible or showed initial promise, rather than to draw firm conclusions about effectiveness. The use of comparison groups was closely linked to the location and design of studies. Studies carried out in online or laboratory settings were much more likely to include clear comparisons between groups, often using random assignment. In contrast, studies conducted in schools or community settings sometimes employed less controlled designs due to practical or ethical constraints. However, many still included some form of comparison, such as an alternative instructional approach. Overall, the heavy reliance on comparative designs underscores a field-wide focus on assessing whether interventions work. In contrast, the continued use of non-comparative designs underscores the importance of exploratory research in developing and refining intervention strategies before they are tested more rigorously. Data collection methods Within the 132 studies included in this scoping review, quantitative data collection methods clearly predominated, reflecting the strongly experimental nature of the intervention literature. The vast majority of studies (around 83%) relied primarily on self-administered questionnaires or structured surveys to assess outcomes. These surveys were most often delivered online and were used to measure changes in participants’ knowledge, attitudes, beliefs, perceptions, or intended behaviours following exposure to an intervention. In many of these studies, survey-based outcome measures were supplemented with experimental process measures, such as manipulation checks, attention checks, or recall questions, to confirm that participants had engaged with the intervention content as intended. While these measures played an important role in validating experimental exposure, they were rarely used as primary outcomes and instead served a supporting function within survey-based designs. A smaller proportion of studies (approximately 13%) incorporated behavioural or performance-based measures. These included tasks designed to assess skills such as information evaluation, decision-making accuracy, or simulated behavioural responses within controlled experimental environments. Although less common than self-report measures, behavioural outcomes were more frequently used in studies explicitly focused on skill development or applied competencies rather than on attitudes or intentions alone. Mixed-methods approaches were relatively uncommon, accounting for fewer than one in ten studies (approximately 9%). When qualitative data were included, they typically took the form of open-ended survey questions, focus groups, or semi-structured interviews conducted alongside quantitative assessments. These qualitative components were primarily used to explore participants’ experiences with the intervention, to assess feasibility and acceptability, or to provide additional context for interpreting quantitative results. Only a minimal number of studies (fewer than 5%) relied exclusively on qualitative data collection methods, such as interviews or observations. These studies were largely exploratory and were more often conducted in educational or community settings than in online experimental contexts. 23 Overall, the data collection methods across the evidence base were dominated by self-reported quantitative measures, with comparatively limited use of behavioural, qualitative, or longitudinal data. This pattern reflects a strong focus on short-term, individual-level outcomes and highlights opportunities for future research to adopt more diverse and ecologically valid approaches to assessing intervention effects. Sample size characteristics Sample sizes in the studies included in our scoping review varied widely: total sample sizes ranged from fewer than 30 participants to more than 1,000, although extensive studies were relatively uncommon. The most common sample size category was moderate-sized studies, with approximately 65 studies (around 50%) enrolling between 100 and 500 participants. This range was particularly characteristic of online experimental studies, in which recruitment via digital platforms or research panels facilitated access to larger samples. A further approximately 20 studies (around 15%) included more than 500 participants, and a small number of these exceeded 1,000 participants. These larger studies were almost exclusively conducted online or within large educational cohorts. At the lower end of the distribution, approximately 35 studies (26%) involved fewer than 100 participants. These smaller-sample studies were disproportionately conducted in educational, community-based, or applied settings, where recruitment was constrained by class size, institutional access, or ethical considerations. Small samples were also more common in studies focusing on specific or vulnerable populations, for whom large-scale recruitment posed additional challenges. Such studies frequently adopted pilot, feasibility, or quasi-experimental designs, rather than fully powered effectiveness evaluations. Where intervention group sizes were reported, a similar pattern of variability was observed. In studies employing a single intervention and a control group, intervention arms typically included 50-200 participants. In studies with multiple intervention conditions, the number of participants per condition was often smaller, sometimes fewer than 50 per group, even when the total sample size exceeded several hundred. Only a small minority of studies (fewer than 15%) explicitly reported a priori sample size calculations or power analyses, and very few provided sufficient information on follow-up participation to allow assessment of attrition or longer-term retention. As a result, most studies can be characterised as short-term evaluations, with sample sizes driven primarily by feasibility or convenience rather than by the requirements of longitudinal outcome assessment or population representativeness. Overall, the distribution of sample sizes reflects a strong reliance on moderate-sized, cross-sectional experimental samples, alongside a smaller number of large-scale online studies and a substantial subset of small applied or exploratory investigations. 24 Population characteristics Across the 132 studies included in this scoping review, study populations varied, but there was a pronounced reliance on convenience and panel-based sampling. University and college students were the most frequently studied population, followed by secondary school students; together, student samples accounted for approximately half of the evidence base, with university students comprising the clear majority. These samples were typically recruited through educational institutions or university-affiliated online platforms and were characterised by relatively homogeneous educational backgrounds. A further one-third of studies drew on general-population samples, commonly recruited via online survey panels or crowdsourcing platforms, with minimal stratification by socioeconomic status, educational attainment, or other key demographic characteristics. Although these samples covered broader age ranges than student populations, they remained largely restricted to individuals with regular internet access, limiting their representativeness. Only a small minority of studies (23; less than one-fifth) examined specific or defined population groups beyond students or general adult samples despite the relevance of such groups to intervention impact and equity. These included children and adolescents recruited through schools, migrant or refugee populations, ethnic or racial minority groups, and individuals with lower levels of education or digital or health literacy. Such studies were more likely to be conducted in school, community, or applied settings and to position these groups as the intended beneficiaries of the interventions. In contrast, fewer than 10 studies focused on professional or occupational groups, such as teachers, educators, or community workers. even in these cases, interventions predominantly emphasised capacity-building among professionals, with limited attention to downstream effects on end users. This pattern underscores a persistent underrepresentation of both structurally vulnerable populations and the practitioners who mediate intervention delivery. Overall, the population profiles of the included studies reveal a heavy reliance on student samples and general-population internet users, with comparatively few studies explicitly designed for distinct or structurally disadvantaged groups. While this pattern likely reflects the relative ease and low cost of recruiting these populations for experimental research, it also exposes significant gaps in the evidence base regarding interventions tailored to more diverse, marginalised, or hard-to-reach communities. With respect to vulnerability, fewer than half of the studies (approximately 50–55) explicitly identified or targeted a disadvantaged population. Where vulnerability was addressed, it usually concerned children and adolescents, ethnic or racial minority groups, individuals with lower levels of education or digital literacy, migrants or refugees, or populations experiencing structural or informational disadvantage. Notably, vulnerability was often treated implicitly rather than being explicitly conceptualised or theoretically grounded, emerging from sample characteristics rather than constituting a central design principle or outcome focus of the intervention. This limits the extent to which the existing literature can inform equitable, context-sensitive intervention development. 25 5.2. Intervention types, scope, and aims: distinguishing AI literacy from other literacy interventions The interventions identified in this scoping review fall into two main groups: those focused on AI literacy and those targeting other forms of literacy, such as media, information, and misinformation literacy. All of these interventions share a common goal of helping people better understand and navigate today’s complex information environments. However, they differ in their focus and in what they are designed to teach. AI literacy interventions focus on helping people understand how artificial intelligence systems work and how they influence information and decision-making. In contrast, other literacy interventions focus more on evaluating information sources, recognising misleading or false content, and developing critical thinking skills when engaging with media and online information. AI literacy interventions AI literacy interventions were specifically targeted at individuals’ understanding of artificial intelligence systems and the algorithmic processes involved in the production, distribution, and personalisation of information. These interventions are typically aimed at demystifying how AI-driven technologies, such as recommender systems, automated content generation tools, and algorithmic ranking mechanisms, shape information exposure and influence visibility, credibility, and trust. The main aims of AI literacy interventions are to enhance conceptual understanding of AI, including basic principles of machine learning, data-driven decision-making, and the limitations and biases embedded in automated systems. Rather than focusing primarily on the veracity of individual content items, these interventions sought to foster system-level awareness, enabling participants to critically reflect on how information is curated, amplified, or suppressed by technological infrastructures. Secondary aims of AI literacy interventions often include promoting critical awareness of ethical, societal, and political implications of AI, such as bias, transparency, accountability, and power asymmetries between platforms and users. In educational contexts, AI literacy was also framed as a future-oriented competence, supporting learners in developing informed attitudes toward emerging technologies and responsible engagement with AI-mediated information ecosystems. Overall, AI literacy interventions emphasised structural and technological literacy, positioning misinformation as one possible consequence of broader algorithmic dynamics rather than as the sole or primary object of intervention. Other literacy interventions (media, information, and misinformation literacy) In contrast, most of the interventions focused on media, information, or misinformation literacy rather than on AI specifically. These interventions aimed to help people evaluate information more carefully by strengthening skills such as assessing the credibility of sources, identifying misleading or false claims, recognising emotionally manipulative content, and thinking critically about news and online information. The primary goal of these interventions was typically to reduce susceptibility to misinformation and improve people's ability to determine what information is accurate or trustworthy. 32 Number of participants reached Participant reach varied widely across the reviewed studies: small and medium sample sizes were most common, particularly in experimental and pilot interventions. Approximately half of the studies reported fewer than one hundred participants, often drawing on single classrooms, specific student cohorts, or controlled experimental samples. These studies were typically designed to assess feasibility, short-term learning outcomes, or specific theoretical mechanisms. At the same time, a substantial minority of studies involved larger participant groups, with samples ranging from several hundred to more than one thousand individuals. These larger-scale interventions were more often linked to online delivery formats, self-directed learning modules, or integration into existing educational programmes, which enabled broader reach without substantial increases in human resource requirements. Only a small number of studies have reported large-scale dissemination, such as nationwide initiatives or platform-based interventions that reach several thousand participants. Overall, the distribution indicates that the field remains dominated by smallto medium-scale implementations, with relatively few examples of sustained or large-scale deployment. Number of sessions delivered The number of sessions delivered closely reflected broader patterns in intervention intensity. Singlesession formats were the most frequently reported, accounting for approximately two-fifths of studies, particularly those employing experimental or quasi-experimental designs. These interventions were commonly delivered as standalone workshops, individual lessons, or single online modules. Multi-session delivery was also widespread. About one-third of studies reported interventions consisting of two to four sessions, typically delivered over a short period, such as one or two weeks. This format was widespread in schooland university-based educational settings. A smaller proportion of studies described more extended delivery models, involving multiple sessions spread over several weeks or months. These longer interventions were often embedded within curricula or structured programmes and were more likely to report cumulative learning outcomes rather than effects associated with a single exposure. Products and materials developed In addition to outcomes measured at the participant level, many studies reported the creation of concrete educational products. These outputs included digital learning modules, lesson plans, instructional videos, interactive tools, serious games, and educational toolkits. Digital products were the most common, appearing in well over half of the studies, and were often designed to be reusable or adaptable beyond the original research setting. Some interventions produced standalone tools or platforms, such as online games or web-based learning environments, that were presented as scalable outputs with potential for broader dissemination. Other studies produced supporting materials, including worksheets, guidelines, or explanatory resources, mainly intended for use by educators or facilitators. Despite the frequency of product development, relatively few studies explicitly reported what happened after the study ended. Information on sustained use, wider uptake, or institutional adoption was limited. In many cases, educational products were described as outputs created for the study, rather than as elements of longer-term implementation or dissemination strategies. 33 Table 5: Overview of Intervention Outputs and Reach (N=132) Output dimension Dominant pattern Indicative prevalence Participants reached Small–medium samples (<100 participants) ~50% of studies Medium–large samples (100–1,000 participants) ~35–40% of studies Very large-scale reach (>1,000 participants) Small minority Sessions delivered Single-session delivery ~40–45% of studies Short multi-session (2–4 sessions) ~30–35% of studies Extended delivery (weeks/months) ~15–20% of studies Products developed Digital educational materials (modules, videos, tools) >50% of studies Standalone tools (e.g. games, platforms) ~15–20% of studies Supporting materials (worksheets, guides, lesson plans) Common Post-intervention uptake Reported reuse or broader dissemination Limited Study-specific or pilot outputs only Majority 5.6 Intervention outcomes Across the 132 studies included in this scoping review, intervention outcomes were evaluated using a combination of measured indicators, such as objective tests or task-based assessments, and selfreported indicators, including perceptual, attitudinal, and confidence-based measures. Although both approaches were widely used, apparent differences emerged in their relative prevalence, the types of outcomes they captured, and their overall robustness, as well as between AI literacy and digital or media literacy interventions. In general, measured outcomes were used more frequently to assess knowledge acquisition and skill development, whereas self-reported outcomes predominated in assessments of attitudes, confidence, and perceived competence. Across both outcome types, evidence of longer-term effects or behavioural change was limited, with few studies examining outcomes beyond the immediate or short-term post-intervention period. Measured outcomes Measured or task-based evaluations, including objective assessments and experimental responses, were reported in approximately 62% of the studies. These outcomes were particularly prominent in digital and media literacy interventions, in which researchers frequently assessed participants’ ability to evaluate information quality, identify misinformation, and apply critical reasoning strategies. Within this group, approximately 55% of studies employed objective measures of information evaluation or critical thinking, such as accuracy scores, misinformation-detection rates, or structuredreasoning tasks. Interventions targeting misinformation and usually grounded in inoculation theory were particularly likely to report measures of resilience to misinformation, typically operationalised as reduced belief in false claims or increased resistance to manipulative techniques under controlled experimental conditions. However, several of these interventions also reported a common unintended effect, namely reduced trust not only in misinformation but also in credible news sources. This spillover effect raises substantive concerns about whether such interventions enhance discernment between false and well-sourced information or instead promote a more generalised scepticism. Attitudinal and behavioural outcomes were otherwise assessed infrequently. Only around 18% of studies included delayed or follow-up measurements, and fewer than 10% examined actual behavioural change, such as news engagement on social media or information-sharing practices, indicating a strong emphasis on short-term cognitive effects. In AI literacy interventions, outcomes were similarly narrow in scope, most often focusing on conceptual understanding of AI systems, including knowledge of algorithms, data use, and automation 34 processes. Approximately 52% of studies employed objective knowledge assessments, whereas outcomes related to technical proficiency, such as programming skills or hands-on interaction with AI models, appeared in fewer than 10% of studies. Overall, this pattern reflects a prevailing focus on declarative knowledge and experimentally induced cognitive resistance, with limited attention to applied skills, behavioural transfer, or potential unintended consequences. Self-reported outcomes Self-reported outcomes were used in approximately 72% of the studies, often alongside measured indicators, but in some cases as the sole form of evaluation. These outcomes were prevalent in assessments of attitudes, confidence, awareness, and perceived skills. In digital and media literacy interventions, self-reported critical thinking, confidence in evaluating information, and perceived media literacy were reported in around 45% of studies. Although many studies documented positive changes in these areas, the heavy reliance on self-assessment raises questions about the extent to which perceived competence aligns with actual skills. Self-reported attitudinal outcomes, including changes in trust toward media or increased caution in information sharing, appeared in approximately 28% of studies and showed greater variability and less consistently positive effects than knowledgeor skill-based outcomes. In AI literacy interventions, self-reported measures played an especially prominent role. About 60% of AI literacy studies assessed perceived understanding of AI systems, awareness of algorithmic bias, or ethical sensitivity toward AI. Attitudinal outcomes, such as increased scepticism toward AI-generated content or diminished trust in automated systems, were typically measured via self-report and appeared in approximately 30% of AI literacy studies. Notably, several studies defined positive outcomes not as increased trust in AI, but as heightened critical awareness and more cautious engagement, reflecting the normative and reflective orientation of many AI literacy interventions. Comparative patterns and methodological gaps Across both AI literacy and digital or media literacy interventions, measured outcomes were most commonly used to assess knowledge and skills, whereas self-reported outcomes predominated in attitudinal, reflective, and awareness-based domains. Despite this clear division, relatively few studies explicitly examined the relationship between objectively measured performance and participants’ perceived competence. As a result, important questions remain regarding calibration, potential overconfidence, and the extent to which self-reported gains reflect actual learning. This issue is particularly pronounced in AI literacy interventions, where the heavy reliance on self-reported outcomes, combined with the limited use of performance-based assessments, highlights a significant methodological gap, especially given the abstract and conceptually complex nature of AI-related knowledge. These challenges are further compounded by the diagnostic complexity of assessing AI-related digital competences. The rapid evolution of AI technologies, the continual emergence of new tools and applications, and the expanding range of possible assessment indicators make it difficult to establish stable, standardised, and comparable measurement frameworks. Consequently, conducting comparative studies across national contexts and over more extended time periods remains methodologically demanding, limiting the field’s capacity to track learning outcomes consistently or to assess long-term trends. 35 Outcomes of digital and media literacy interventions Outcomes related to digital and media literacy knowledge and skills were the most frequently reported across the reviewed studies. Approximately 65% of studies documented improvements in participants’ information-evaluation abilities, including enhanced source assessment, more accurate judgments of information quality, and greater sensitivity to misleading or manipulative cues. Closely related outcomes in critical thinking and epistemic vigilance were also prominent, with around 50% of studies reporting gains in analytical reasoning, reflective judgment, or scepticism toward unverified information. Notably, these outcomes were often framed as improvements in reasoning processes rather than as changes in specific beliefs About 35% of studies explicitly examined resilience to misinformation, often focusing on healthrelated content and younger adults. Reported outcomes included reduced belief in false claims, increased resistance to manipulation, and improved detection of misinformation, with some studies showing these effects could persist even in the presence of further corrective information. Interventions grounded in inoculation theory were particularly likely to produce such effects, though measurements were mostly immediate, and attitudinal outcomes, such as changes in media trust or news engagement practices, were less consistently reported. Sustained behavioural or belief changes were rarely assessed. More general media literacy outcomes, such as confidence in navigating digital environments, were reported in roughly one-third of studies, typically as secondary outcomes. An overview of the outcome types can be found in Figure 2 below. This graph includes only 120 studies; 12 of 132 had missing data on this item. Figure 2: Outcome overview Overall, the evidence highlights the complexity of media literacy interventions and the conditional nature of their effects. Many interventions risk negative spillover on trust in accurate information. The problem is that discerning trustworthy from untrustworthy information is inherently difficult, given the complexity of the information environment and the varied forms of biased or problematic content audiences encounter (Hameleers, 2024). Current interventions that focus on “fake” or misleading news primarily influence subjective responses to misinformation, while offering little guidance on how to identify trustworthy sources reliably. Preexisting media trust appears to play a key role, suggesting 36 that tailoring messages to existing levels of (dis)trust may be critical for effectiveness (Hameleers, 2024). Outcomes of AI literacy interventions AI literacy interventions displayed a partially distinct outcome profile compared with digital and media literacy interventions. The most consistently reported outcomes concerned understanding of AI concepts and systems, with approximately 63% of AI literacy studies documenting improvements in conceptual knowledge related to algorithms, data-driven decision-making, or AI-generated content. These gains were largely conceptual rather than technical, prioritising comprehension and critical understanding over operational or programming skills. A second prominent outcome domain concerned awareness of the biases, limitations, and ethical implications of AI systems. Around 45% of AI literacy interventions reported increased recognition of algorithmic bias, system opacity, or broader societal risks associated with AI technologies. These outcomes were frequently framed in normative or critical terms, emphasising participants’ capacity to question the assumed neutrality, authority, or objectivity of automated systems. Outcomes related to interaction with AI tools or applications, such as familiarity with AI-enabled technologies or improved interpretation of AI outputs, were reported in a smaller subset of studies, accounting for approximately 23% of interventions. Assessments of advanced technical competencies, such as programming or model development, were rare and confined to a small subset of studies. Attitudinal outcomes in AI literacy interventions were mixed and less consistently reported. Approximately 28% of studies documented shifts in trust, scepticism, or perceived agency regarding AI systems. Notably, these changes did not uniformly reflect increased trust; instead, many studies framed positive outcomes as the development of greater critical awareness and more cautious engagement with AI-mediated information. Comparative patterns and outcome gaps Comparative analysis indicates that digital and media literacy interventions primarily targeted content-level evaluation, reasoning skills, and resistance to misinformation. In contrast, AI literacy interventions placed greater emphasis on system-level understanding, awareness of bias, and ethical reflection. Across both categories, knowledgeand skill-based outcomes were reported more consistently than attitudinal or behavioural outcomes. Evidence for long-term impact, transfer of skills across contexts, or sustained behaviour change remained limited. Fewer than 20% of studies included delayed follow-up measures, and outcome measures were frequently tailored to individual studies, which constrained comparability across the literature. Across the 132 studies, outcomes spanned a wide range of cognitive, skill-based, attitudinal, and epistemic domains. Despite substantial variation in how outcomes were operationalised and when they were assessed, several common patterns emerged. Most interventions reported positive shortterm effects, particularly in knowledge acquisition and skill development, whereas evidence for longer-term or behaviour-oriented outcomes was comparatively sparse. 37 Table 6: Outcome Domains by Intervention Type (N = 132) A. Digital and Media Literacy Outcomes Outcome domain Description Indicative prevalence Predominant assessment Information knowledge Understanding of misinformation, news production, and credibility cues ~60–65% of studies Mostly measured Evaluation skills Source evaluation, accuracy judgments, misinformation detection ~55–60% Mostly measured Critical thinking / epistemic vigilance Analytical reasoning, scepticism, reflective judgment ~45–50% Mixed Resilience to misinformation Resistance to false or manipulative content ~30–40% Mostly measured Attitudes toward media Trust, caution, perceived credibility ~25–30% Mostly self-reported Broader digital/media literacy Perceived competence navigating information environments ~30–35% Mostly self-reported Behavioral outcomes Information sharing, media practices <20% Mixed; rarely longitudinal B. AI Literacy Outcomes Outcome domain Description Indicative prevalence (AI studies) Predominant assessment AI concept understanding Algorithms, data use, automation, AIgenerated content ~60–65% Mixed (measured > self-report) Awareness of AI bias & limitations Bias, opacity, fairness, accountability ~40–50% Mostly self-reported AI ethics & societal implications Power, governance, responsibility ~30–40% Mostly self-reported AI tool awareness/interpretation Familiarity with AI outputs or applications ~20–25% Mostly self-reported Attitudes toward AI Trust, scepticism, perceived agency ~25–30% Mostly self-reported Technical AI skills Programming, model interaction Small minority Measured 5.7 Outcome effectiveness: direction, magnitude, and strength of evidence Among the 132 studies, the efficacy of interventions was assessed employing a diverse combination of quantitative and qualitative methodologies, exhibiting considerable variation in analytical rigour, reportingpractices, and targeted outcomes domains. Overall, the evidence base points to a generally positive pattern of effects, particularly for short-term cognitive and skill-based outcomes. At the same time, the strength, consistency, and comparability of effectiveness evidence varied widely across studies, limiting the extent to which firm conclusions can be drawn. 38 Direction of effects The majority of quantitative studies reported effects in the expected or intended direction. Approximately 75% of studies documented statistically significant improvements in at least one primary outcome following the intervention. These improvements most frequently concerned knowledge acquisition, evaluative skills, and critical thinking, and, in the case of misinformationfocused interventions, reduced susceptibility to misleading or false content. A smaller proportion of studies reported mixed effects, in which some outcomes improved while others did not, or where effects varied across participant subgroups. Null effects were observed in a minority of studies and were most often associated with attitudinal outcomes, delayed follow-up measures, or complex constructs such as behavioural change. Reports of adverse effects were rare. In AI literacy interventions, the overall direction of effects was similarly positive, particularly for conceptual understanding of AI systems and awareness of bias and ethical implications. However, effects on attitudes toward AI, including trust or acceptance, were more heterogeneous. In several studies, positive outcomes were framed not as increased trust but as greater scepticism or more critical engagement with AI systems, underscoring the normative orientation of many AI literacy interventions. Effect size reporting Effect sizes were reported inconsistently across the reviewed studies, and only around one-third of quantitative studies explicitly reported standardised effect sizes, such as Cohen’s d, eta squared, or odds ratios, which substantially limits opportunities for cross-study comparison. When effect sizes were reported, they were most often small to moderate, particularly for short, single-session interventions. Larger effects were more commonly observed in studies employing tightly controlled experimental designs that focused on narrowly defined cognitive outcomes, such as specific knowledge tests, or in studies that implemented interventions with repeated exposure across multiple sessions. By contrast, outcomes related to attitudes, beliefs, or broader literacy constructs generally produced smaller, more variable, or less consistently reported effects. In AI literacy interventions, reported effect sizes were typically modest and concentrated on conceptual knowledge and awareness, rather than on behavioural outcomes or technical skill development. Statistical significance Statistical significance was reported in most quantitative studies, typically based on pre–post comparisons or between-group contrasts. Approximately two-thirds of studies reported statistically significant effects for at least one outcome measure. However, interpretation of these findings is constrained by the heavy reliance on short-term post-intervention assessments, limited reporting of confidence intervals, and infrequent adjustment for multiple comparisons. In several studies, statistically significant effects on primary outcomes were reported alongside nonsignificant findings for secondary outcomes, suggesting selective or domain-specific effectiveness rather than uniform intervention impact. Follow-up analyses assessing the durability of effects were uncommon. Only a small subset of studies reported statistically significant long-term outcomes, further limiting conclusions about the intervention's sustained effectiveness. Quantitative and qualitative evidence Quantitative evidence predominated across the corpus, with most studies relying on survey-based or experimental outcome measures. Qualitative findings, while reported in a smaller but substantively 39 important subset of studies, were typically derived from interviews, open-ended survey responses, or classroom observations. These qualitative insights generally complemented and contextualised quantitative results, shedding light on how and why interventions produced their effects. Participants frequently reported increased awareness, reflection, and more critical engagement with information or AI systems, even when quantitative effect sizes were modest. In AI literacy interventions, qualitative data often revealed shifts in participants’ thinking about AI, including heightened awareness of system limitations, bias, and broader societal implications. At the same time, some qualitative findings indicated tensions or ambivalence, such as heightened awareness of algorithmic bias accompanied by uncertainty or reduced confidence. Standardised quantitative measures did not always capture these nuanced responses. From a scoping perspective, these patterns suggest a field characterised by generally promising but methodologically fragmented evidence of effectiveness. Together, they underscore the need for more standardised outcome reporting, longer-term assessment, and stronger integration of quantitative and qualitative methods to support more robust and cumulative conclusions about intervention impact. Table 7: Overview of Outcome Effectiveness (N = 132) Effectiveness dimension Dominant pattern Indicative prevalence Direction of effects Positive effects on at least one outcome ~70–80% of studies Mixed or domain-specific effects ~15–20% of studies Null or no detectable effects Small minority Effect size reporting Standardised effect sizes reported ~30–35% of studies No effect size reported Majority Magnitude of effects Small to moderate effects Most reported cases Large effects Rare Statistical significance ≥1 statistically significant outcome ~65–70% of studies No statistically significant effects Minority Outcome domains with the strongest effects Knowledge and evaluative skills Common Attitudes and behaviours Less consistent AI literacy–specific effectiveness Improved AI concept understanding and bias awareness Common Effects on trust or acceptance of AI Mixed Evidence type Quantitative only Majority Mixed quantitative–qualitative ~20–25% of studies Qualitative only Small subset 5.8 Challenges and facilitators of implementing misand disinformation/AI literacy interventions Across the included studies, a range of practical, pedagogical, technological, and conceptual factors emerged that either facilitated or constrained the implementation of misand disinformation as well as AI literacy interventions. These factors can be inferred from how interventions were designed, resourced, delivered, and evaluated, as well as from patterns observed in reported outcomes and effectiveness. Facilitators of implementation A central facilitating factor across interventions was the use of low-threshold, accessible resources. Most interventions relied on widely available digital infrastructures, such as standard devices, webbased platforms, and classroom-based settings, and did not require advanced technical expertise. This accessibility-supported implementation was effective across diverse institutional contexts and enhanced the scalability of many interventions, particularly within formal education systems. 40 Human facilitation by educators or researchers also emerged as a key enabler. Facilitated delivery contextualised abstract concepts, supported guided reflection, and helped learners engage with complex topics, including strategies for misinformation and algorithmic bias. This role was especially important in AI literacy interventions, where the higher conceptual complexity required facilitators to bridge the gap between technical systems and learners’ everyday experiences. From a pedagogical perspective, interventions that combined direct instruction with active learning elements, such as guided analysis, scenario-based exercises, or structured discussion, were more consistently associated with positive outcomes. Implementation was further facilitated when interventions were integrated into existing curricula or classroom routines, particularly in multisession formats. At a conceptual level, the use of established theoretical and pedagogical frameworks, including media and information literacy, inoculation theory, and socio-technical perspectives on AI, provided a shared language and structure for intervention design. These frameworks supported coherence between aims, activities, and outcomes, and helped align interventions with broader educational and policy objectives. Finally, the development of reusable educational resources, such as digital modules, lesson plans, or educational games, served as an additional facilitator of implementation by enabling replication and adaptation beyond the original study context, even when long-term adoption was not systematically assessed. Challenges to implementation Despite these facilitating factors, several recurring implementation challenges were evident across the corpus. 1. A first challenge concerned time and intensity constraints. Many interventions were delivered as single sessions or very short multi-session formats, often due to curricular limitations, institutional pressures, or experimental research designs. These constraints restricted the depth of engagement and reduced opportunities for consolidation, transfer across contexts, and longer-term impact. 1. A second challenge related to conceptual complexity, particularly in AI literacy interventions. Communicating algorithmic systems, data processes, and ethical implications in ways that were both accessible and accurate proved demanding. Consequently, many interventions prioritised conceptual understanding and awareness over hands-on technical engagement, thereby limiting opportunities for deeper system interaction or transferable skill development. 2. Uneven theoretical articulation posed a third challenge: a substantial proportion of studies did not clearly specify their theoretical foundations, potentially weakening coherence in intervention design and increasing reliance on local interpretation by facilitators. This lack of explicit theoretical grounding also complicates replication, comparison, and the cumulative development of knowledge across studies. 3. As a fourth challenge, from an evaluative perspective, misalignment between intervention goals and outcome measures emerged as an implicit concern. Although many interventions sought to foster critical thinking, resilience, or ethical awareness, evaluation often relied on short-term or self-reported outcomes, limiting insight into whether these aims translated into durable or behaviourally meaningful change. 41 4. Resource-related challenges were less often technological than human and organisational. While interventions rarely required advanced technical infrastructure, practical implementation, particularly in AI literacy contexts, depended heavily on facilitator confidence and conceptual understanding. Limited training or ongoing support for facilitators may therefore have constrained the quality of implementation in some settings. Finally, scalability and sustainability were weakly addressed. Although many interventions were designed to be scalable in principle, few studies have described systematic strategies for institutional embedding, long-term maintenance, or post-project adoption, indicating an implementation gap between pilot success and sustained practice. Compared with other interventions, misinformation and disinformation literacy interventions benefited from more established pedagogical traditions and clearer instructional routines, which facilitated classroom implementation. AI literacy interventions, while increasingly prominent, faced greater challenges related to abstraction, ethical complexity, and facilitator preparedness, but were also supported by strong normative framing around fairness, accountability, and broader societal relevance. 5. Discussion This scoping review shows a rapidly expanding but methodologically uneven evidence base on interventions addressing misinformation and disinformation, as well as AI literacy, particularly for vulnerable populations. Across 132 studies, interventions were generally feasible to implement, often relying on accessible digital infrastructure and facilitated delivery. Most reported positive short-term outcomes in knowledge and evaluative skills, but evidence for durable impact, behavioural change, and transfer across contexts remains limited. These patterns have direct implications for how PRODIGI designs, evaluates, and scales its intervention work packages. A central methodological implication concerns the dominance of self-reported measures in the field and the risks this creates for interpreting effectiveness. As highlighted in methodological reflections from the REMEDIS project (Tomczyk, 2023), self-assessment is vulnerable to subjective biases shaped by participants’ self-perceptions, expectations, and social desirability, including the common assumption that educational interventions should produce improvement across all indicators. In digital, media, and AI-related competence domains, this is particularly consequential because increased awareness may coexist with unchanged, underestimated, or overestimated actual skills (Heinecke et al., 2019). Moreover, the articles included in our study point to a systematic misalignment between perceived and measured competence in complex cognitive domains such as critical thinking, media evaluation, and algorithmic understanding, with overconfidence among lower-skilled learners and underestimation among more competent participants. These calibration effects complicate comparisons across studies using different self-report instruments and weaken internal validity when self-report is treated as stand-alone evidence of change. Relatedly, many self-assessment tools used in intervention research appear methodologically shallow in our included articles, frequently relying on short ad hoc questionnaires or single-item indicators with limited psychometric validation. Such measures may be susceptible to short-term enthusiasm, motivation, or familiarity with terminology rather than reflecting durable learning outcomes demonstrated through objective tests, tasks, or behavioural indicators (Yates et al., 2022). In the reviewed corpus, this problem is compounded by the scarcity of designs that assess long-term effectiveness. Few studies used delayed post-tests or post-post test designs, and only a small proportion reported long-term effect sizes. As a result, it remains difficult to distinguish immediate cognitive gains from sustained changes in digital competences, media practices, or AI-related 48 8. Appendices A. Code Book, V.2 Group 1: Study_Characteristics 1. Country – Country where the intervention took place 2. Setting_Type – School, University, community, NGO, etc. 3. Study_Design – RCT, quasi-experimental, pre–post, qualitative, mixed-methods, etc. 4. Has_Comparison_Group – Yes/No 5. Data_Collection_Methods – Survey, test, interview, focus group, observation, log data, etc. 6. Sample_Size_Total 7. Sample_Size_Intervention 8. Age_Range – e.g., 12–18 9. Population_Description – e.g., “migrant youth” 10. Vulnerable_Group_Focus – Yes/No (indicate only if there is a vulnerable group) 11. Vulnerable_Group_Type – e.g., unaccompanied minors; socio-economically disadvantaged youth; elderly in rural areas; migrants/refugees; other Group 2: Intervention_LogicModel 1. Intervention_Label_in_Paper – the name used in the article 2. Intervention_Aim_Main – main goal (e.g., “increase resilience to misinformation”) 3. Intervention_Aim_Secondary – e.g., “build AI awareness”, “enhance critical thinking” Group 3: Inputs (Logic Model – what goes in): 1. Inputs_Human_Resources – teachers, trainers, librarians, NGO staff, peer mentors, etc. 2. Inputs_Technology – devices, platforms, AI tools, LMS, social media, etc. 3. Inputs_Materials – curriculum, lesson plans, OER, videos, games, toolkits 4. Inputs_Other – funding, partnerships, physical spaces, etc. Group 4: Activities (Logic Model – what is done): 1. Activities_Description_Short – 2–3 line summary of the key activities 2. Activities_Type – workshop, course, curriculum, social media campaign, etc. 3. Delivery_Mode – face-to-face, online, blended, hybrid 4. Group_Format – individual, small group, class, community group 5. Session_Count – how many sessions 49 6. Session_Duration_Minutes – typical session length 7. Total_Duration_Hours – approximate total dose 8. Intervention_Period_Weeks – total length in weeks/months 9. Facilitation_Model – teacher-led, peer-led, researcher-led, co-design, etc. 10. Theoretical_Frameworks_Reported – media literacy, critical pedagogy, AI literacy framework, no theory, etc. Group 5: Outputs (Logic Model – immediate products): 1. Output_Number_of_Participants_Reached 2. Output_Number_of_Sessions_Delivered 3. Output_Products_Developed – e.g., Open Educational Resources (OERs), teaching toolkit, videos, game, chatbot, campaign materials Group 6: Outcomes (Logic Model – changes observed/measured): 1. Outcomes_Knowledge – e.g., “increase in AI/ ML knowledge test scores” 2. Outcomes_Skills – e.g., “improved search or verification skill” 3. Outcomes_Attitudes – e.g., “reduced trust in unverified sources” 4. Outcomes_Critical_Thinking – e.g., explicit critical thinking outcomes (if measured) 5. Outcomes_Resilience_to_Misinformation – ability to detect or resist mis/disinformation 6. Outcomes_AI_Literacy – understanding AI concepts, tools, or bias 7. Outcomes_Digital_or_Media_Literacy – broader digital or media literacy outcomes 8. Outcomes_Other – e.g., wellbeing, empowerment, civic engagement, feeling positive or negative 9. Outcomes_news_engagement – e.g., discussing the news with family or friends, searching online about the news, sharing news on social media, doing nothing, etc. 10. Followup_Assessment_Conducted – Yes/No 11. Follow-up_Timeframe – e.g., 3 months later Group 7: Outcomes_Effectiveness 1. Outcome_Effect_Size_Reported – effect size(s) if provided (e.g., d = 0.45) 2. Outcome_Statistical_Significance – e.g., p < .05, NS 3. Key_Quantitative_Results_Summary – 2–3 lines summarising main numeric findings 4. Key_Qualitative_Results_Summary – themes like “participants report feeling more critical toward news sources” 50 5. Challenges– Barriers or any reported difficulties in terms of intervention design, implementation, and assessment 6. Facilitators– Any reported enablers or facilitators in terms of intervention design, implementation, and assessment 7. Recommendations– Any reported recommendations and suggestions for different stakeholders in the fields, e.g., for policy makers, teachers, librarians, NGOs, etc. 51 B. Overview of included studies Row Author(s) Year Title Journal DOI 1. Ali & Qazi 2023 Countering misinformation on social media through educational interventions: Evidence from a randomized experiment in Pakistan Journal Of Development Economics 10.1016/j.jd eveco.2023. 103108 2. Ali et al. 2022 Fake news on Facebook: examining the impact of heuristic cues on perceived credibility and sharing intention Internet Research 10.1108/IN TR-10-20190442 3. Alon et al. 2024 Fighting fake news on social media: a comparative evaluation of digital literacy interventions: Research and Reviews Current Psychology 10.1007/s1 2144-02405668-4 4. Alsaad & AlDossary 2024 Educational Video Intervention to Improve Health Misinformation Identification on WhatsApp Among Saudi Arabian Population: Pre-Post Intervention Study Jmir Formative Research 10.2196/50 211 5. Appel et al. 2025 Psychological inoculation improves resilience to and reduces willingness to share vaccine misinformation Scientific Reports (Nature Publisher Group) 6. Apuke & Gever 2023 A quasi experiment on how the field of librarianship can help in combating fake news Journal Of Academic Librarianship 10.1016/j.a calib.2022.1 02616 7. Apuke et al. 2023 Effect of Fake News Awareness as an Intervention Strategy for Motivating News Verification Behaviour Among Social Media Users in Nigeria: A Quasi-Experimental Research Journal Of Asian And African Studies 10.1177/00 219096221 079320 8. Apuke et al. 2023 The effect of visual multimedia instructions against fake news spread: A quasi-experimental study with Nigerian students Journal Of Librarianship And Information Science 10.1177/09 610006221 096477 9. Apuke et al. 2023 Literacy Concepts as an Intervention Strategy for Improving Fake News Knowledge, Detection Skills, and Curtailing the Tendency to Share Fake News in Nigeria Child And Youth Services 10.1080/01 45935X.202 1.2024758 10. Armeen et al. 2025 Combating Fake News Using Implementation Intentions Information Systems Frontiers 10.1007/s1 0796-02410502-0 11. Artmann et al. 2023 Elementary school students' information literacy: Instructional design and evaluation of a pilot training focused on misinformation Journal Of Media Literacy Education 10.23860/J MLE-202315-2-3 12. Aruguete et al. 2024 Framing fact-checks as a “confirmation” increases engagement with corrections of misinformation: a four-country study Scientific Reports 10.1038/s4 1598-02453337-0 13. Axelsson et al. 2024 Bad News in the civics classroom: How serious gameplay fosters teenagers’ ability to discern misinformation techniques Journal Of Research On Technology In Education 10.1080/15 391523.202 4.2338451 14. Badrinathan 2021 Educative Interventions to Combat Misinformation: Evidence from a Field Experiment in India The American Political Science Review 10.1017/S0 003055421 000459 15. Batista Pereira et al. 2023 Inoculation Reduces Misinformation: Experimental Evidence from Multidimensional Interventions in Brazil Journal Of Experimental Political Science 10.1017/XP S.2023.11 16. Berger et al. 2025 Debunking “fake news” on social media: Immediate and short-term effects of fact-checking and media literacy interventions Journal Of Public Economics 10.1016/j.jp ubeco.2025 .105345 52 17. Bowles et al. 2025 Sustaining Exposure to Fact-Checks: Misinformation Discernment, Media Consumption, and Its Political Implications American Political Science Review 10.1017/S0 003055424 001394 18. Bruns et al. 2024 Investigating the role of source and source trust in prebunks and debunks of misinformation in online experiments across four EU countries Scientific Reports 10.1038/s4 1598-02471599-6 19. Burel et al. 2024 Exploring the impact of automated correction of misinformation in social media Ai Magazine 10.1002/aa ai.12180 20. Capecchi et al. 2024 A Gamified Platform to Support Educational Activities about Fake News in Social Media Ieee Transactions On Learning Technologies 21. Capewell et al. 2024 Misinformation interventions decay rapidly without an immediate posttest Journal Of Applied Social Psychology 10.1111/jas p.13049 22. Costello et al. 2024 Durably reducing conspiracy beliefs through dialogues with AI Science 10.1126/sci ence.adq18 14 23. Craig & Vijaykumar 2023 One Dose Is Not Enough: The Beneficial Effect of Corrective COVID-19 Information Is Diminished If Followed by Misinformation Social Media + Society 10.1177/20 563051231 161298 24. Dame AdjinTettey 2022 Combating fake news, disinformation, and misinformation: Experimental evidence for media literacy education Cogent Arts And Humanities 10.1080/23 311983.202 2.2037229 25. Dobber et al. 2025 Rhyme or reason: medium-term effects of heuristic and traditional media literacy interventions Information Communication & Society 10.1080/13 69118X.202 5.2533315 26. Domgaard & Park 2021 Combating misinformation: The effects of infographics in verifying false vaccine news The Health Education Journal 10.1177/00 178969211 038750 27. Dutta-Powell et al. 2024 Two interventions for mitigating the harms of greenwashing on consumer perceptions Business Strategy And The Environment 10.1002/bs e.3520 28. Ecker et al. 2024 Don’t believe them! Reducing misinformation influence through source discreditation: Principles and Implications Cognitive Research 29. Ecker et al. 2024 Don't believe them! Reducing misinformation influence through source discreditation Cognitive ResearchPrinciples And Implications 10.1186/s4 1235-02400581-7 30. Erişen et al. 2024 Exploring the effectiveness of virtual reality in combating misinformation on climate change Political Psychology 10.1111/po ps.13057 31. Ferrucci & Hopp 2023 Let’s intervene: How platforms can combine media literacy and self-efficacy to fight fake news Communication And The Public 10.1177/20 570473231 203081 32. Ford et al. 2023 Online Media Literacy Intervention in Indonesia Reduces Misinformation Sharing Intention Journal Of Media Literacy Education 33. Gambin & Munzel 2025 Guarding against deception through news literacy interventions: Impact of issue involvement and participation on young French adults’ fake news vulnerability New Media And Society 10.1177/14 614448251 342796 34. Georgiou et al. 2025 Effectiveness of the Scientific Reasoning Intervention on Reducing Online Conspiracy Beliefs and Misinformation Engagement: A Study Using the (Mis)Information Game Applied Cognitive Psychology 10.1002/ac p.70069 35. Green et al. 2022 Active versus passive: evaluating the effectiveness of inoculation techniques in relation to misinformation about climate change Australian Journal Of Psychology 10.1080/00 049530.202 2.2113340 53 36. Gross & Balaban 2025 The Effectiveness of an Educational Intervention on Countering Disinformation Moderated by Intellectual Humility Media And Communication 10.17645/m ac.9109 37. Gruener 2021 Identifying and Debunking EnvironmentallyRelated False News Stories - An Experimental Investigation German Journal Of Agricultural Economics 10.30430/gj ae.2021.01 76 38. Guo et al. 2025 Specific media literacy tips improve AI-generated visual misinformation discernment Cognitive Research: Principles And Implications 10.1186/s4 1235-02500648-z 39. Hameleers 2022 Separating truth from lies: comparing the effects of news media literacy interventions and factcheckers in response to political misinformation in the US and Netherlands Information, Communication & Society 10.1080/13 69118X.202 0.1764603 40. Hameleers 2024 Is the alarm on deception ringing too loudly? The effects of different forms of misinformation warnings on risk perceptions of misinformation exposure European Journal Of Communication 10.1177/02 673231241 271015 41. Hameleers 2022 Populist Disinformation: Are Citizens With Populist Attitudes Affected Most by Radical Right-Wing Disinformation? Media And Communication 10.17645/m ac.v10i4.56 54 42. Hameleers & van der Meer 2023 Striking the balance between fake and real: under what conditions can media literacy messages that warn about misinformation maintain trust in accurate information? Behaviour And Information Technology 10.1080/01 44929X.202 3.2267700 43. Harley et al. 2025 Evaluating an animated, story-driven media literacy video to improve media credibility accuracy and understand the role of epistemic emotions in credibility ratings European Journal Of Psychology Of Education 10.1007/s1 0212-02500969-z 44. Harrop et al. 2023 Inoculation Can Reduce the Perceived Reliability of Polarizing Social Media Content International Journal Of Communication 45. Hassan & Sabry 2024 Leveraging the ability of the online health information seekers to find credible online sources The Egyptian Journal Of Community Medicine (Egypt) 10.21608/ej cm.2024.24 9600.1276 46. Heley et al. 2025 Mitigating Health and Science Misinformation: A Scoping Review of Literature from 2017 to 2022 Health Communication 10.1080/10 410236.202 4.2332817 47. Helfers & Ebersbach 2023 The differential effects of a governmental debunking campaign concerning COVID-19 vaccination misinformation Journal Of Communication In Healthcare 10.1080/17 538068.202 2.2047497 48. Hoes et al. 2024 Prominent misinformation interventions reduce misperceptions but increase scepticism Nature Human Behaviour 10.1038/s4 1562-02401884-x 49. Hruschka & Appel 2023 Learning about informal fallacies and the detection of fake news: An experimental intervention Plos One 10.1371/jou rnal.pone.0 283238 50. Hu et al. 2023 Game-based inoculation versus graphic-based inoculation to combat misinformation: a randomized controlled trial Cognitive ResearchPrinciples And Implications 10.1186/s4 1235-02300505-x 51. Huang et al. 2024 Media Literacy Interventions Improve Resilience to Misinformation: A Meta-Analytic Investigation of Overall Effect and Moderating Factors Communication Research 52. Iles et al. 2021 Investigating the Potential of Inoculation Messages and Self-Affirmation in Reducing the Effects of Health Misinformation Science Communication 10.1177/10 755470211 048480 54 53. Jalbert et al. 2025 Social Truth Queries: Development of a New UserDriven Intervention for Countering Online Misinformation Journal Of Applied Research In Memory And Cognition 10.1037/ma c0000142 54. Janmohamed et al. 2021 Interventions to Mitigate COVID-19 Misinformation: A Systematic Review and MetaAnalysis Journal Of Health Communication 10.1080/10 810730.202 1.2021460 55. Jiang et al. 2022 Inoculation works and health advocacy backfires: Building resistance to COVID-19 vaccine misinformation in a low political trust context Frontiers In Psychology 10.3389/fps yg.2022.976 091 56. Johnson & Sparks 2024 Narrative misinformation from a credible source can be discredited with counternarrative Jcom-Journal Of Science Communication 10.22323/2. 23080202 57. König et al. 2022 The Development and Evaluation of an e-Learning Course That Promotes Digital Health Literacy in School-age Children: Pre-Post Measurement Study Journal Of Medical Internet Research 10.2196/37 523 58. Kops et al. 2025 Young people and false information: A scoping review of responses, influential factors, consequences, and prevention programs Computers In Human Behavior 10.1016/j.c hb.2025.10 8650 59. Kozyreva et al. 2024 Toolbox of individual-level interventions against online misinformation Nature Human Behaviour 10.1038/s4 1562-02401881-0 60. Ku et al. 2025 Helping young students cope with the threat of fake news: efficacy of news literacy training for junior-secondary school students in Hong Kong Educational Studies 10.1080/03 055698.202 3.2296345 61. Le et al. 2025 Building confidence in the digital age: the effects of online professional development on Vietnamese teachers’ media and information literacy Asia Pacific Journal Of Education 10.1080/02 188791.202 5.2505670 62. Leder et al. 2024 Feedback Exercises Boost Discernment of Misinformation for Gamified Inoculation Interventions Journal Of Experimental Psychology-General 10.1037/xg e0001603 63. Lee & Bissell 2024 User agency-based versus machine agency-based misinformation interventions: The effects of commenting and AI fact-checking labeling on attitudes toward the COVID-19 vaccination New Media & Society 10.1177/14 614448231 163228 64. Lee et al. 2024 Building resilience to misinformation in communities of color: Results from two studies of tailored digital media literacy interventions New Media And Society 10.1177/14 614448241 227841 65. List et al. 2024 Critical thinking and misinformation vulnerability: experimental evidence from Colombia Pnas Nexus 10.1093/pn asnexus/pg ae361 66. Lu et al. 2023 Psychological Inoculation for Credibility Assessment, Sharing Intention, and Discernment of Misinformation: Systematic Review and MetaAnalysis Journal Of Medical Internet Research 10.2196/49 255 67. Ma et al. 2023 Fighting COVID-19 Misinformation through an Online Game Based on the Inoculation Theory: Analyzing the Mediating Effects of Perceived Threat and Persuasion Knowledge International Journal Of Environmental Research And Public Health 10.3390/ijer ph2002098 0 68. Marten & Stadtler 2025 Building resilience against online misinformation: A teacher-led training promoting evaluation strategies among lower secondary students Computers In Human Behavior 10.1016/j.c hb.2024.10 8548 55 69. Marten et al. 2025 Did 5G radiation really kill birds? Training lower secondary students in epistemic strategies to counter online misinformation Learning And Individual Differences 10.1016/j.li ndif.2025.1 02685 70. McGrew 2024 Teaching lateral reading: Interventions to help people read like fact checkers Current Opinion In Psychology 10.1016/j.c opsyc.2023. 101737 71. McGrew & Chinoy 2022 Fighting misinformation in college: students learn to search and evaluate online information through flexible modules Information And Learning Science 10.1108/ILS -09-20210081 72. Mirhoseini et al. 2023 Actively open-minded thinking is key to combating fake news: A multimethod study Information & Management 10.1016/j.i m.2023.103 761 73. ModirroustaGalian et al. 2023 Effects of Inductive Learning and Gamification on News Veracity Discernment Journal Of Experimental Psychology-Applied 10.1037/xa p0000458 74. Moore & Hancock 2022 A digital media literacy intervention for older adults improves resilience to fake news Scientific Reports 10.1038/s4 1598-02208437-0 75. Naderer & Opree 2021 Increasing Advertising Literacy to Unveil Disinformation in Green Advertising Environmental Communication-A Journal Of Nature And Culture 10.1080/17 524032.202 1.1919171 76. Neylan et al. 2023 How to “inoculate” against multimodal misinformation: A conceptual replication of Roozenbeek and van der Linden (2020) Scientific Reports 10.1038/s4 1598-02343885-2 77. Nygren & Efimova 2025 Investigating the long-term impact of misinformation interventions in upper secondary education Plos One 10.1371/jou rnal.pone.0 326928 78. Okuhara et al. 2025 Effectiveness and determinants of narrative-based corrections for health misinformation: A systematic review Patient Education And Counseling 10.1016/j.p ec.2025.109 253 79. Oliveira et al. 2024 Confronting misinformation related to health and the environment: a systematic review Journal Of Science Communication 10.22323/2. 23010901 80. Oliveira Moreira et al. 2024 JEDi - a digital educational game to support student training in identifying portuguese-written fake news: Case studies in high school, undergraduate and graduate scenarios Education And Information Technologies 10.1007/s1 0639-02312309-z 81. Ono & Ogata 2025 Framework of a Disinformation Game Based on the Narrative Warfare in Russo-Ukrainian War Journal Of Robotics, Networking And Artificial Life 10.57417/jr nal.11.1_11 82. Opgenhaffen 2022 Fact-Checking Interventions on Social Media Using Cartoon Figures: Lessons Learned from the Tooties Digital Journalism 10.1080/21 670811.202 1.2011758 83. Orosz et al. 2023 A prosocial fake news intervention with durable effects Scientific Reports 10.1038/s4 1598-02330867-7 84. Orosz et al. 2024 Strategies to combat misinformation: Enduring effects of a 15-minute online intervention on critical-thinking adolescents Computers In Human Behavior 10.1016/j.c hb.2024.10 8338 85. Orosz et al. 2024 Softly empowering a prosocial expert in the family: lasting effects of a counter-misinformation intervention in an informational autocracy Scientific Reports 10.1038/s4 1598-02461232-x 56 86. Panizza et al. 2022 Lateral reading and monetary incentives to spot disinformation about science Scientific Reports 10.1038/s4 1598-02209168-y 87. Paula Gil 2024 Gamification as a Methodology to Enhance Analytical and Sustainable Engagement on Social Media Discover Education 10.1007/s4 4217-02300074-7 88. Peng et al. 2024 The media literacy dilemma: can ChatGPT facilitate the discernment of online health misinformation? Frontiers In Communication 10.3389/fco mm.2024.1 487213 89. Pillai & Fazio 2023 Explaining Why Headlines Are True or False Reduces Intentions to Share False Information Collabra-Psychology 10.1525/col labra.87617 90. Porter et al. 2024 Shoves, nudges and combating misinformation: evidence on a new approach Behavioural Public Policy 10.1017/bp p.2024.42 91. Pretus et al. 2024 The Misleading count: an identity-based intervention to counter partisan misinformation sharing Philosophical Transactions Of The Royal Society BBiological Sciences 10.1098/rst b.2023.004 0 92. Qian et al. 2023 Fighting cheapfakes: Using a digital media literacy intervention to motivate reverse search of out-ofcontext visual misinformation Journal Of ComputerMediated Communication 10.1093/jc mc/zmac02 4 93. Rasmussen et al. 2024 Public Health Communication Reduces COVID-19 Misinformation Sharing and Boosts Self-Efficacy Journal Of Experimental Political Science 10.1017/XP S.2024.2 94. Rau & Premo 2025 Systematic Review of Educational Approaches to Misinformation Educational Psychology Review 10.1007/s1 0648-02510012-8 95. Redaelli et al. 2025 Mastering critical thinking skills is strongly associated with the ability to recognize fakeness and misinformation Frontiers In Education 10.3389/fe duc.2025.1 577692 96. Reinhardt et al. 2025 Experimental testing of active vs. passive inoculation interventions on adolescents’ news discernment and learning enjoyment Journal Of Media Psychology: Theories, Methods, And Applications 97. Rho et al. 2025 Inoculating students against misleading data visualizations with perceptual training and informative feedback Computers & Education 10.1016/j.c ompedu.20 25.105413 98. Roozenbeek et al. 2022 Technique-based inoculation against real-world misinformation Royal Society Open Science 10.1098/rso s.211719 99. Roozenbeek et al. 2022 Psychological inoculation improves resilience against misinformation on social media Science Advances 10.1126/sci adv.abo625 4 100. Roozenbeek et al. 2021 Disentangling Item and Testing Effects in Inoculation Research on Online Misinformation: Solomon Revisited Educational And Psychological Measurement 10.1177/00 131644209 40378 101. Skipper et al. 2023 'But wait, that isn't real': A proof-of-concept study evaluating 'Project Real', a co-created intervention that helps young people to spot fake news online British Journal Of Developmental Psychology 10.1111/bjd p.12456 102. Smith et al. 2025 Inoculation reduces social media engagement with affectively polarized content in the UK and US. Communications Psychology 10.1038/s4 4271-02500189-7 103. Spearing et al. 2025 Countering AI-generated misinformation with preemptive source discreditation and debunking Royal Society Open Science 10.1098/rso s.242148 104. Stein et al. 2023 Realtime user ratings as a strategy for combatting misinformation: an experimental study Scientific Reports 57 105. SwireThompson et al. 2024 Discrediting Health Disinformation Sources: Advantages of Highlighting Low Expertise Journal Of Experimental Psychology-General 10.1037/xg e0001627 106. Taerning et al. 2023 A tale about facts and opinions: The impact of a drama intervention on middle-school students' information literacy Interaction Design And Architectures 107. Tamboer et al. 2024 Do You Know What Fake News Is? An Exploration of and Intervention to Increase Youth’s Fake News Literacy Youth And Society 10.1177/00 44118X231 205930 108. Tang et al. 2025 Effects of a News Literacy Video on News Literacy Perceptions and Misinformation Evaluation Media And Communication 10.17645/m ac.8983 109. Tay et al. 2022 A comparison of prebunking and debunking interventions for implied versus explicit misinformation British Journal Of Psychology 10.1111/bjo p.12551 110. Taylor et al. 2025 Mapping the Landscape of K-12 Interventions That Target Students' Evaluation of Science-Related Misinformation or Disinformation School Science And Mathematics 10.1111/ss m.18365 111. Tessa L. 2024 Breaking 'Fake News': Attempting to Mitigate the Spread of False Information through Media Literacy Education in Adolescents Using the Inoculation Theory Proquest LLC 112. Thomas et al. 2022 Pilot study suggests online media literacy programming reduces belief in false news in Indonesia First Monday 113. Tseng et al. 2021 Fighting “bad science” in the information age: The effects of an intervention to stimulate evaluation and critique of false scientific claims Journal Of Research In Science Teaching 10.1002/tea .21696 114. Udry & Barber 2025 The effects of inoculation interventions and repetition on perceived truth in younger and older adults. Cognition 10.1016/j.c ognition.20 25.106275 115. van Erkel et al. 2024 Combating Disinformation With News Literacy Interventions: An Experimental Study on the Framing Effects of News Literacy Messages International Journal Of Press/Politics 10.1177/19 401612241 279534 116. Velez et al. 2023 Latino-Targeted Misinformation and the Power of Factual Corrections Journal Of Politics 10.1086/72 2345 117. Verhalle & Loos 2025 Fighting Disinformation: How to Measure the Impact of Preand Debunking on Dutch Primary School Children’s Media Literacy? Societies 10.3390/so c15060155 118. Viguer & Picó 2025 Development and validation of an intervention protocol to promote media and information literacy for critical thinking in adolescents’ use of social media Social Sciences And Humanities Open 10.1016/j.ss aho.2025.1 01807 119. Vijaykumar et al. 2021 How shades of truth and age affect responses to COVID-19 (Mis)information: randomized survey experiment among WhatsApp users in UK and Brazil Humanities And Social Sciences Communications 10.1057/s4 1599-02100752-7 120. Walter et al. 2021 Evaluating the Impact of Attempts to Correct Health Misinformation on Social Media: A MetaAnalysis Health Communication 10.1080/10 410236.202 0.1794553 121. Wedlake et al. 2024 How games can support misinformation education: A sociocultural perspective Journal Of The Association For Information Science And Technology 10.1002/asi .24954