Dataset from the article "LLMs and the Illusion of Rigor: implications on Global Asymmetry and AI Governance in the International System"

Quote 2 Quote Analysis [2] 1st Quote: ●Statement in the text: "...two-level game analysis, pioneered by Robert Putnam 2;" ●Verification: The source is the 1988 article "Diplomacy and domestic politics: the logic of two-level games" by Robert D. Putnam. The work is recognized as the pioneering article that introduced the concept of "two-level game analysis" into international relations theory. ●Result: CA (Proper Citation). The attribution is accurate and the source is the seminal academic work on the subject. 2nd Citation: ●Statement in the text: "This mapping reveals that AI governance is not a single, monolithic negotiation but a fluid and contested field of interaction where domestic politics (Level II) and international diplomacy (Level I) are inextricably linked.2" ●Verification: Putnam's central argument is that international politics (Level I) and domestic politics (Level II) are interlinked. The article explores how this link works. The statement in the text applies this logic to the field of AI governance, constituting a theoretically consistent use of the source. ●Result: CA (Appropriate Citation). The source supports the logic that diplomacy (Level I) and domestic politics (Level II) are "inextricably linked". 3rd Citation: ●In-text assertion: "As two-level game logic suggests, no government can make a credible commitment at the international level (Level I) that does not have sufficient support at the domestic level (Level II).2" ●Verification: Putnam devotes a section to the idea of "ratification", arguing that agreements at Level I need to be ratified at Level II. Failure to obtain domestic support leads to "involuntary defection", which corresponds to the statement. ●Result: CA (Appropriate Citation). The statement is an accurate summary of one of the mechanisms described in Putnam's theory. 4th Citation: ●Statement in the text: "By applying the two-level game logic central to this paper's framework, we can deconstruct how the internal nested games-the bargaining between government agencies, powerful industries, and other domestic stakeholders (Level II)- directly shape the international negotiating positions and strategies (Level I) of each actor.2" ●Verification: Putnam describes the actors at Level II (domestic) as including party figures, parliamentarians and interest group representatives. The assertion that bargaining between these actors shapes international positions (Level I) is the article's main thesis. ●Result: CA (Appropriate Citation). The source provides the theoretical basis for the statement, which describes the central dynamic of the model. 5th Citation: ●Statement in the text: "The two-level nested game framework effectively explained how the distinct international strategies of the United States, the European Union, and China are direct reflections of their internal political bargains.2" ●Verification: Putnam's article uses the example of the 1978 Bonn Summit to show how the international strategies of different countries were reflections of their internal political bargains. The statement in the text applies this same logical principle to his study. ●Result: CA (Appropriate Citation). The use of the source to support the central logic of the argument is appropriate. Why was this source used? The usage pattern of citation [2] is similar to that of citation [1] and follows a standard academic approach. ●Theoretical Rationale: Robert Putnam's seminal work on two-level game theory was identified. The five occurrences use the source to establish and reinforce the theoretical pillars of the article. ●Consistent Use: No factual distortions or inaccuracies were observed. Each citation refers to a central and specific aspect of Putnam's theory, demonstrating a consistent application of the logical structure of an academic argument. ●The source is used to build the foundation on which the article's argument about AI Governance is developed. Quote 3 Citation Analysis [3] 1st Citation (single occurrence): ●Statement in the text: "The analytical framework combining nested games ¹, two-level games ³, and bureaucratic politics ⁴ provides the necessary tools to map this ecosystem, identifying the key" ●Verification: The statement uses the concept of "two-level games" as one of the pillars of the framework. The source provided is the abstract of the article "Diplomacy and domestic politics: the logic of two-level games" by Robert Putnam. The abstract states: "this article offers a theory of ratification This theory of 'two-level games' may also be applicable to many other political phenomena". Therefore, the source is the appropriate reference for the theory mentioned. ●Result: CA (Proper Citation). The source is correct and supports the use of the concept o f "two-level games" in the theoretical framework. Why was this source used? This case presents a procedural pattern related to the management of sources, and not to the content of the citation itself. ●Duplicate Source: Citation [3] refers to the same work as Citation [2]: Robert Putnam's 1988 article. The only difference identified is the URL from which the source was extracted. ●Lack of Deduplication: The same article was located at multiple addresses and, instead of being consolidated into a single reference, generated two separate entries in the bibliography. Consequently, two different citation numbers (2 and 3) were applied in the text to refer to the same theory. ●Conclusion: Although the citation is classified as "Proper" (CA) because the content of the source is correct, the act of using Citation [3] represents a procedural redundancy. This results in a duplicate and unnecessary reference in the bibliography and text. Quote 4 Citation Analysis [4] 1st Citation: ●Statement in the text: "...bureaucratic politics model, articulated by Graham Allison and Philip Zelikow.⁴" ●Verification: The source, an article by Christopher M. Jones, states that the model was introduced by Graham Allison in 1969 and 1971, and that Philip Zelikow only co-authored the second edition of Essence of Decision (1999). Therefore, attributing the original model to both of them is inaccurate. The source cited addresses the issue, but the statement it supports in the text contains this inaccurate attribution. ●Result: CI (Imprecise Citation). The source is on topic, but the statement contains an authorial inaccuracy that is not drawn from the source itself. 2nd Citation: ●Statement in the text: "...concepts from bureaucratic politics help explain how policy outcomes are often not the product of a single rational choice but are "political resultants" emerging from the clash of competing interests within and between institutions.⁴" ●Verification: The source defines the bureaucratic policy model in a similar way, stating that "foreign policy decisions are the product of political resultants or bargaining between individual leaders in government positions". ●Result: CA (Adequate Citation). The secondary source defines the concept of "political resultants" precisely. 3rd Citation: ●In-text assertion: "By applying this synthesized framework, we can map the key "Players in Position," their preference hierarchies, their institutional bargaining capacities, and the "Action-Channels" through which they exert influenceconcepts drawn directly from Allison and Zelikow's work.⁴" ●Verification: The source mentions "players in positions" and defines "Action-channels" as central concepts of Allison's model. The citation supports the origin of the terms. ●Result: CA (Proper Citation). The source confirms that the technical terms are part of the bureaucratic policy model. 4th Citation: ●Statement in the text: "The analytical framework combining nested games ¹, two-level games ³, and bureaucratic politics ⁴ provides the necessary tools to map this ecosystem, identifying the key" ●Verification: The text states that the "bureaucratic politics" model is one of the three theoretical pillars. The source is an encyclopedic article that defines this model. ●Result: CA (Adequate Citation). The citation of a secondary source is used to substantiate the inclusion of a theoretical term. 5th Citation: ●In-text assertion: "The concept of "Players in Position," drawn from Allison and Zelikow's bureaucratic politics model, requires an assessment not only of who the actors are but also of their hierarchical standing, their institutional constraints, and the capacities they bring to the bargaining table.⁴" ●Verification: The source confirms "Players in Position" as a concept of the model. The description of what the concept "requires" is an elaboration consistent with the theory, although it is not a direct quote from the source. ●Result: CA (Adequate Citation). The source anchors the technical term, and the text provides an interpretation of what it means in practice. 6th Citation: ●Statement in the text: "The negotiation of global AI governance norms primarily occurs within a few key international organizations, which function as the "arenas" or "Action-Channels" where the interests of the various players collide.⁴" ●Verification: The text applies the term "Action-Channels" to the context of international organizations. The source defines "Action-Channels" as "regularized means of taking action on a specific type of governmental issue". The use of the source to justify the application of the term to a real case is an academic practice. ●Result: CA (Proper Citation). The source is used to substantiate the application of a theoretical concept. 7th Citation: ●In-text assertion: "These forums, or "Action-Channels" in the language of Allison and Zelikow, are not neutral platforms; their membership, mandates, and internal rules profoundly shape the negotiations that take place within them.⁴" ●Verification: The statement uses "Action-Channels". The source supports the origin of the term and the idea that these channels have "formal or informal rules that govern political interactions", which validates the statement that they "shape the negotiations". ●Result: CA (Adequate Citation). The source supports both the origin of the term and the description of its function. 8th Citation: ●In-text assertion: "The analysis confirms that the emergent global AI governance regime is not a rationally designed system for the optimal good, but a "political resultant" forged in the crucible of competing interests and profound structural asymmetries.⁴" ●Verification: The statement uses the term "political resultant" in the conclusion of the article. The source defines the concept, and the usage is consistent with the terminology established in the text. ●Result: CA (Appropriate Citation). The use of the theoretical term is consistent with the conclusions of the analysis. Why was this source used? The citation usage pattern [4] indicates a specific citation strategy. ●Source for Definitions: Instead of citing Allison's primary work in each instance, a single secondary source (an encyclopedia article) was used as a reference to define and justify the technical terms associated with the theory of bureaucratic politics. ●Terminological Consistency: The citation [4] is used consistently whenever a specific technical term from the model ("political outcome", "Players in Position", "Action-Channels") is introduced or applied. This practice indicates to the reader the origin of the terminology. ●Additional Note: The only factual inaccuracy observed was the attribution of the original model to "Allison and Zelikow", since the second edition of the main work has both as authors. Apart from this inconsistency, the source was used eight times to build one of the article's theoretical pillars. Quote 5 Citation Analysis [5] 1st Citation (only occurrence): ●Statement in the text: "Consequently, the international frameworks that emerge are "political resultants": suboptimal compromises that bear the imprint of the power dynamics and political bargains that produced them.⁵" ●Verification: The statement defines the technical term "political resultants". The source provided is a summary of Hayden J. Smith's article on the bureaucratic politics model. Although Smith's article discusses the model generally, the specific term "political resultants" does not appear in the summary provided. The source addresses the general topic, but does not contain the specific terminology being defined and cited. ●Result: CI (Imprecise Citation). The statement uses a specific technical term that is not present in the cited source. The source is thematically relevant, but does not contain the terminology used in the statement. Why was this source used? This case presents a different procedural pattern to the previous ones, characterized by thematic association and redundancy. ●Thematic association: Source [5], an article on the bureaucratic politics model, was cited in a sentence that uses the technical term "political resultants". Although the concept belongs to this theory, the specific term was not found in the cited source. The citation seems to have been applied on the basis of a general topic correspondence ("bureaucratic politics"), rather than a specific content correspondence. ●Citation redundancy: Citation [4], used elsewhere in the text, had already been used to define the term "political resultants". The use of Citation [5] in this instance is therefore redundant and introduces an imprecision, since the source [5] does not contain the term in question. ●Conclusion: The use of Citation [5] demonstrates a pattern in which a source is applied to a specific claim on the basis of a thematic overlap, without the exact content of the source supporting the claim. Quote 6 Citation Analysis [6] 1st Citation: ●In-text assertion: "This methodological approach, which has been effectively applied to deconstruct complex foreign policy decisions such as the normalization of U.S.-Cuba relations, offers a powerful lens through which to analyze the emerging field of global Al governance." ●Verification: The statement indicates that the article's methodology (combining Tsebelis, Putnam and Allison) has already been used to analyze the normalization of U.S.-Cuba relations. The source is Guevara Martinez's thesis, which explicitly states in its theoretical chapter the intention to combine the theory of "nested games" (Tsebelis), the logic of "two-level games" (Putnam) and concepts from Allison and Zelikow's "Model III". ●Result: CA (Appropriate Citation). The source is a direct example of the application of the methodology that the article proposes to use, validating the claim that this approach has already been applied to a foreign policy case. 2nd Citation: ●Claim in the text: "A clear hierarchy emerges in the Al governance landscape, distinguishing between a small number of powerful "Chiefs" who set the agenda and a vast number of "Indians" who must navigate a field largely defined by others." ●Verification: The statement uses the terminology "Chiefs" and "Indians" to describe a hierarchy of power. The source, Guevara Martinez's thesis, in discussing Allison and Zelikow's concepts, uses and explains this terminology to differentiate between players at the top and bottom of a decision-making hierarchy. ●Result: CA (Proper Citation). The source is cited to justify the use of specific theoretical jargon. Why was this source used? The use of this source demonstrates a research strategy that goes beyond primary theoretical sources. ●Source as Methodological Precedent: In addition to citing the primary sources of the theories, the text references a recent academic work (a doctoral thesis) that had already employed the same methodological combination. O The use of this thesis acts as a precedent, indicating that the proposed approach has already been applied in another study. ●Source for Conceptual Extraction: The thesis is cited both to validate the general methodology and to introduce specific jargon ("Chiefs" and "Indians") that it had already applied from the original sources. There is the use of a secondary source to extract concepts in a consolidated way. ●In-text assertion: "However, this push for strong regulation was met with intense counter-pressure from the technology industry and some EU member states who expressed concerns that overly strict regulation could stifle Europe's nascent AI industry.¹¹" ●Verification: The source discusses this concern, mentioning that critics maintain that a strict regulatory regime could deter companies and lead to "technological stagnation". ●Result: CA (Adequate Citation). The source directly corroborates the existence of this counter-pressure and concern. Why was this source used? The citation usage pattern [11] indicates that a single academic article was used as the basis for an extended section of the argument. The pattern mixes content extraction with the addition of external information. ●Extraction and Distortion: While some statements are summaries of the source content, others deviate from it. Cases of: ○Inferences and Generalizations: Broader conclusions are drawn from specific source data, extrapolating its scope (e.g. Citation 2, Citation 6). ○Insertion of External Jargon: Well-known technical terms (e.g., "digital sovereignty", "Brussels effect") are introduced into the text and attributed to the source, even though they do not appear in it (e.g., Citation 4, Citation 7). ●Conclusion: This pattern results in a situation where the authority of an academic source is used to support assertions that contain information external to its content, mixing faithful extraction with generalizations and the addition of unquoted concepts. Citation 12 Citation Analysis [12] 1st Quote: ●In-text assertion: "The EU's strict regulations, for example, can be seen as an attempt to create a competitive advantage for European firms in "trustworthy AI".¹²" ●Verification: The source discusses the EU's strategy, mentioning that it is based on "promoting ethical outcomes through the protection of fundamental rights", and also references the "Ethics Guidelines for Trustworthy AI" and the search for "competitiveness". The statement that the regulation seeks to create a competitive advantage through "trustworthy AI" is a summary of the source's argument. ●Result: CA (Appropriate Citation). The source connects the EU's regulatory and ethical approach with its competitiveness goals. 2nd Citation: ●In-text assertion: "The EU AI Act is the quintessential "political resultant" of a complex, multi-level nested game within the EU's unique institutional structure, reflecting a deepseated preference for protecting fundamental rights.¹²" ●Verification: The source discusses the AI Act and the EU's preference for "protecting fundamental rights". However, the source does not use the theoretical terminology "political resultant" or "nested game". The text applies concepts from sources [1] and [4] to the topic of source [12] and cites source [12] as if it contained this analysis. ●Result: CD (Distorted Citation). The text carries out a theoretical analysis, but incorrectly attributes it to this source, which does not contain such a framework. 3rd Citation: ●Statement in the text: "In international forums, the EU champions its rights-based, human-centric approach, positioning it as the gold standard for trustworthy AI.¹²" ●Verification: The source states that in EU documents there is a focus on fostering "human-centric" AI and discusses the EU's power to set standards. The statement is a description of the EU's stance, as detailed in the source. ●Result: CA (Adequate Citation). The source supports the description of the EU's approach. 4th Citation (in Table 1): ●In-text assertion: "Exports its regulatory model (the ""Brussels Effect""). Aims for the AI Act to become the de facto global standard. Champions a human-centric, rights-based approach.¹²" ●Verification: The source supports the description of a "human-centered and rightsbased" approach and the idea of becoming a "global standard". However, the source does not mention the term "Brussels Effect". ●Result: CD (Distorted Citation). Known external jargon is inserted and attributed to a source that does not contain it. Why was this font used? The pattern of use of this source indicates an inconsistent attribution of theoretical analysis. ●Combination of Sources in Citation: The source [12], which describes the EU's approach to AI, is used as a reference for statements that incorporate theoretical concepts (e.g., "political outcome", "nested game") and jargon (e.g., "Brussels effect") that are not present in its text. These concepts and jargon are taken from other sources or from external knowledge. ●Conclusion: This method of citation creates the impression that the source [12] contains a specific theoretical analysis (using the lenses of "nested games" and "policy outcomes"), which does not correspond factually to its content. Quote 13 This is our third case of "orphan source": the reference [195] is in the bibliography, but was not used in the body of the article. Citation Analysis [ ]14 1st Citation (only occurrence): ●Claim in text: "[The Big Tech lobby's] primary preference is for a light-touch, voluntary regulatory environment that avoids high compliance costs and protects their market advantages.¹⁴" ●Verification: The statement describes the US regulatory preference as "light-touch" and voluntary. The source, the abstract of Zhang and Lyu's article, states that the US adopts a model that emphasizes free development, stimulating market vitality through "industry self-discipline and flexible regulation". The description in the source corresponds to a "light and voluntary" environment, and the objective of "stimulating market vitality" aligns with "protecting market advantages". ●Result: CA (Adequate Citation). The source corroborates the characterization of the regulatory approach mentioned in the statement. Why was this source used? The use of this source demonstrates a pattern of redundancy, similar to that observed in other analyses. ●Conceptual redundancy: The description of the US regulatory approach, the subject of this quote, had already been comprehensively covered by the source [11]. A new academic source was introduced to support a single sentence on a topic already covered. ●Citation pattern: The introduction of a new reference for an already established point, rather than the reuse of the previous source, indicates a pattern of collecting multiple sources on the same topic that are distributed throughout the text rather than being consolidated. Although the content of the citation is correct, its inclusion is procedurally redundant. Citation Analysis [ ]15 1st Quote: ●In-text assertion: "While not a monolithic bloc, countries across Africa, Latin America, and Asia share common challenges, including significant resource disparities and limited influence in global standard-setting bodies.¹⁵" ●Verification: The source is a press release about an event focused on Latin America and the Caribbean, not mentioning Africa or Asia. The source talks about "high inequality", which aligns with "resource disparities", but does not discuss "limited influence in global standard-setting bodies". ●Result: CIR (Citation Irrelevant). The source is specific to one region, while the statement is a broad statement about three continents, with elements not contained in the source. 2nd Citation: ●Statement in the text: "This infrastructure gap creates a fundamental dependency on technology and cloud services provided by corporations from the Global North.¹⁵" ●Verification: The press release does not mention "infrastructure gap", "dependency", "cloud services" or "corporations from the Global North". ●Result: CI (Imprecise Citation). The source addresses the general theme of AI challenges in the region, but does not contain the specific information about technological or infrastructure dependency. 3rd Citation: ●Statement in the text: "This concentration of resources creates a formidable barrier to entry for actors from the Global South, who are often forced to react to frameworks and standards set elsewhere.¹⁵" ●Verification: The concepts of "concentration of resources" and "barrier to entry" are not mentioned in the source. Although the source quotes a line about the need f o r a governance model of its own for Latin America, it does not state that actors are "forced to react". ●Result: CI (Imprecise Citation). The source does not contain a specific statement about barriers to entry or the obligation to react to external standards. 4th Quote: ●In-text assertion: "The AI landscape in Latin America, as detailed in reports from the Economic Commission for Latin America and the Caribbean (ECLAC) , is more heterogeneous.¹⁵" ●Verification: The source, an ECLAC press release, quotes an authority who says: "...where there is some kind of regulation, it varies from country to country". This variation supports the idea that the scenario is "heterogeneous". ●Result: CA (Appropriate Citation). The source contains a statement that validates the claim that the AI landscape in the region is varied. 5th Citation: ●Statement in the text: "The prevailing approach is one of strategic adaptation, where countries adopt and tailor international principles from forums like the OECD and UNESCO to their local contexts while fostering regional cooperation.¹⁵" ●Verification: The source does not mention "strategic adaptation", "OECD" or "UNESCO". It mentions the need for "more coordination between countries", but does not describe the strategy of adapting principles from specific forums. ●Result: CI (Imprecise Citation). The source supports the idea of regional cooperation, but not the premise about adapting principles from specific forums. 6th Citation (in Table 2): ●Statement in the text: "Strategic Adaptation: Adopting and tailoring international principles (OECD, UNESCO) to regional contexts.¹⁵" ●Verification: The verification is identical to the previous one; the source does not mention the terms or the strategy. ●Result: CI (Imprecise Citation). Why was this source used? The pattern of use of this citation is characterized by the application of a source of limited scope to support analytical generalizations. ●Quality and Appropriateness of Source: The source used is a press release about an event. This type of document does not have the depth of a research report and was used to support complex and wide-ranging analytical arguments, for which it is not suitable. ●Generalization from Content: Only one of the six statements (the 4th) is directly supported by the content of the source. The other five make geographical generalizations (Quote 1) or introduce technical and geopolitical concepts (Quotes 2, 3 and 5) not mentioned in the press release. The pattern observed consists of making detailed statements based on a source that is only thematically related, but whose content does not specifically support them. Quote 16 Analysis of Quote [16] 1st Quote: ●In-text assertion: "Through regional bodies like the African Union (AU), which has developed its own Continental AI Strategy, these nations are attempting to build collective bargaining power and articulate an alternative vision for AI governance centered on inclusive development and digital sovereignty.¹⁶" ●Verification: The source is the official document "Continental Artificial Intelligence Strategy" of the African Union. The document emphasizes an "Africa-centric" and "inclusive development-oriented" approach. It also highlights the need to "strengthen African participation in global AI governance" and develop "sovereign AI capabilities", which aligns with the concepts of "collective bargaining power" and "digital sovereignty". ●Score: CA (Adequate Citation). The statement is an accurate summary of the core objectives of the strategy described in the source. 2nd Citation: ●In-text statement: "The African Union's Continental AI Strategy is a comprehensive example of a regional effort to build collective bargaining power.¹⁶" ●Verification: The document emphasizes "regional cooperation", "intra-African coordination" and the need to develop "common African positions", which are components for building "collective bargaining power". ●Result: CA (Appropriate Citation). The source is a direct example of the effort mentioned in the statement. 3rd Citation (in Table 2): ●In-text assertion: "AU Continental AI Strategy (2024) ¹⁶" ●Verification: The title of the source is "Continental Artificial Intelligence Strategy" and the publication date is "July 2024". ●Result: CA (Proper Citation). The title and year are correct. 4th Citation: ●Statement in the text: "Continued investment in regional coalition-building, exemplified by the AU's Continental AI Strategy, is the most effective means of increasing bargaining power.¹⁶" ●Verification: The African Union document is a continental strategy that aims to unify member states around a common vision for AI. Its pillars, such as "Regional and International Cooperation", make it an example of "regional coalition building". ●Result: CA (Adequate Citation). The source serves as an example for the statement made. Why was this source used? The usage pattern of this quote demonstrates the application of a primary source to the construction of a central argument. ●Use of Primary Source: The primary, official policy document on the topic was located and used, rather than relying on secondary sources. ●Conceptual Synthesis: The content of the strategic document was synthesized to extract high-level objectives (inclusive development, collective bargaining power, sovereignty), summarizing them concisely. ●Integration into the Argument: The source was integrated into the article's argument, serving as the main example of a regional strategy for the Global South, which was the focus of that section. Quote 17 Quote Analysis [17] 1st Quote: ●In-text statement: "Adopted in 2024, the strategy outlines an "Africa-centric" approach rooted in African values and explicitly prioritizes inclusive growth, local talent development, and digital sovereignty.¹⁷" ●Verification: The statement describes the Continental AI "Strategy" (source [16]), but appends the quote [17], which refers to the African "Declaration" on AI. Although the content of source [17] is thematically similar, the quote is attributed to the wrong document. Additionally, source [17] describes an event dated April 2025, while the statement in the text mentions adoption in 2024, creating a time inconsistency. ●Result: CD (Distorted Citation). An attribution error occurs, as the statement describes the content of one document (the Strategy [16]), but cites another (the Declaration [17]), confusing the two sources . 2nd Citation (in Table 2): ●Statement in the text: "Africa-centric Development: Prioritizing local needs and African values. Capacity Building: Investing in local talent and data infrastructure. Collective Bargaining: Harmonizing policies to create a unified African position.¹⁷" ●Verification: This is a description of the priorities of the AI Continental Strategy (source [16]). The source cited [17] (the Declaration) supports these themes, mentioning reflection on "strategic priorities, shared values" and the "development of AI talent". However, the attribution of the document remains incorrect. ●Result: CD (Distorted Citation). Similar to the previous citation, the text describes the Strategy [16], but attributes the citation to the Declaration [17], mixing up the sources. Why was this source used? The use of this quote presents multiple factual and procedural inconsistencies. ●Conflict between sources: Confusion is observed between two documents on AI policy in Africa: the "Continental AI Strategy" (source [16]) and the "African Declaration on AI" (source [17]). The text describes the content of the first document, but attributes the citation to the second. ●Quality and Temporal Anomaly of the Source: The source [17] is an article from the website of a legal association, not an official document. Analysis of its content reveals an inconsistency: the article describes an event dated April 2025, a future date in relation to the creation date of the text analyzed (July 2024). Quote 21 Citation Analysis [21] 1st Citation (single occurrence): ●In-text claim: "Organizations like Access Now, the AI Now Institute, and the Future of Life Institute conduct critical research on the societal impacts of AI, campaign against harmful applications like biometric surveillance, and lobby for stronger protections in legislation like the EU AI Act.²¹" ●Verification: The statement generalizes the work of three organizations. The source cited is a report published by Access Now, which confirms the activities described for Access Now itself. However, the document does not mention the "AI Now Institute", the "Future of Life Institute" or the topic of "biometric surveillance". The source can only validate the claim in relation to itself. ●Result: CI (Imprecise Citation). A "citation bundling" occurs. The claim is about three entities, but the source provided, belonging to only one of them, is used to validate the generalization. The source does not substantiate the claims about the other two organizations. Why was this source used? The usage pattern in this quote is characterized by "clustering", where a source of limited scope is generalized. ●Partial Validation: The source, an Access Now report, was identified to prove the organization's work in the area of AI. Using the organization's own report to describe its activities is a straightforward approach. ●Source generalization: Inaccuracy occurs when extending the validity of this quote to cover the other two organizations mentioned in the same sentence. A single source, belonging to one of the actors, has been applied to a statement that encompasses three different actors Quote 22 This is another case of "orphan source": the reference [206] is in the bibliography, but was not actually used in the body of the article. Quote 23 This is another case of "orphan source": the reference [207] is in the bibliography, but was not actually used in the body of the article. Quote 24 This is our seventh case of "orphan source": the reference [208] is in the bibliography, but was not actually used in the body of the article. Citation analysis [25] 1st Quote (only occurrence): ● Statement in the text: "Technical standards bodies, most notably the Institute of Electrical and Electronics Engineers (IEEE) and the International Organization for Standardization (ISO), play a vital role in developing the granular, operational standards for AI safety, transparency, and ethics.²⁵" ● Verification: The statement refers to the role of two organizations, IEEE and ISO. The source provided is an IEEE document that confirms the activities described for the organization itself, including the development of standards for "transparency" and "safety". However, the document does not mention ISO. ● Result: CI (Imprecise Citation). The pattern of "citation clustering", observed in the citation [21], is repeated. The source is used to validate an assertion about two entities, but its content only supports the part referring to IEEE. Why was this source used? The pattern of "citation clustering" is observed again. ● Partial Validation: An official IEEE document was identified to describe its role as a standardization body in AI. ● Scope of Source: Inaccuracy occurs when using this single source to validate the entire sentence, which also includes ISO. The quote is inaccurate because the scope of the source does not support the entirety of the statement. Citation 26 This is another case of "orphan source": the reference [210] is in the bibliography, but was not actually used in the body of the article. It is also the second source on the same IEEE initiative that has been collected but not used or poorly referenced. Citation Analysis [27] 1st Citation: ● Statement in the text: "The OECD, a forum of predominantly high-income, marketbased economies... producing the influential OECD AI Principles.²⁷" ● Result: CI (Imprecise Citation) / CQB (Low Quality Citation). The statement is inaccurate because it adds a geopolitical characterization that is not in the source. In addition, the source is an informational web page, a format considered to be of low robustness for an academic article . 2nd Citation: ● Statement in the text: "The OECD AI Principles, first adopted in 2019 and updated in 2024, are a landmark achievement and the first intergovernmental standard on AI.²⁷" ● Verification: The source contains these facts. However, the nature of the source (informational web page) makes it not very robust to the context. ● Result: CQB (Low Quality Citation). The information is correct, but the source is not academic. 3rd Citation (in Table 3): ● Statement in the text: "Non-binding intergovernmental standard.²⁷" ● Result: CI (Imprecise Citation) / CQB (Low Quality Citation). The citation is inaccurate because it adds the term "non-binding", which is not in the source. The fact that it is an informational web page contributes to the low quality rating. Why was this source used? The pattern of use of this source indicates a selection that does not distinguish between "official" and "academic" sources. ● Source Type Selection: The source selected is the official OECD website which describes its own principles. There is a prioritization of a primary source of information over sources of academic analysis on the same topic. Although the source is official, its format (publicity page) does not correspond to that of a scientific publication, making it less robust for the context. The standard suggests prioritizing the origin of the information over the suitability of the source's format for academic work. Citation Analysis [28] 1st Citation: ● In-text assertion: "It [the United States] champions the adoption of high-level principles, such as those in the OECD framework, which it helped shape, but resists calls for legally binding international treaties.²⁸" ● Verification: The source is an OECD webpage about its own principles and does not discuss the specific diplomatic position of the United States in relation to them. ● Result: CI (Imprecise Citation) / CQB (Low Quality Citation). The citation is inaccurate because the source does not contain information about the US position. In addition, the source is not academic. 2nd Citation: ● In-text assertion: "The principles are deliberately high-level and non-binding, promoting values such as "inclusive growth," "human-centred values," and "transparency" while recommending that governments create an "enabling policy environment".²⁸" ● Verification: The source lists the values of "inclusive growth", "transparency" and the recommendation of an "enabling governance and policy environment", as well as a "human-centered approach". However, the terms "high-level" and "non-binding" are not present in the source text. ● Result: CI (Imprecise Citation) / CQB (Low Quality Citation). The information extracted from the source is combined with external details ("high-level", "non-binding") that are not in the quoted text. The source is also of low quality for the context. 3rd Quote (in Table 3): ● In-text statement: "Innovation, trustworthy AI, human rights, democratic values, transparency, accountability.²⁸" ● Verification: The source lists principles that correspond to the terms of the statement, such as "Human rights and democratic values", "Transparency" and "Accountability", and mentions the promotion of "innovative and trustworthy" AI. ● Result: CQB (Low Quality Citation). Although the list of values is accurate, the use of an informative web page as a source for a table in an academic article is a practice of low robustness. Why was this source used? The use of this source demonstrates patterns of redundancy and source type inappropriateness. ● Source redundancy: The citation [28] refers to an OECD web page covering the same topic as the citation [27]. Two different URLs on the same subject have been treated as two separate sources, a procedural error that leads to duplicate references. ● Citation Patterns: The pattern of using informative sources from official websites as if they were academic sources continues, without discerning the type of evidence appropriate to the context. for the context. Additionally, the pattern of adding details that are not contained in the cited source (e.g., "non-binding") is also repeated, resulting in inaccuracies. Citation 29 Citation Analysis [29] 1st Quote: ● Statement in the text: "The UN Secretary-General's High-level Advisory Body on AI and UNESCO's Recommendation on the Ethics of AI reflect a broader, more developmentoriented perspective on governance.²⁹" ● Verification: The source states that UNESCO's standards aim to contribute to a "more inclusive, sustainable and peaceful world" and lists "inclusive growth, sustainable development and well-being" as one of its values, which supports the assertion of a "development-oriented perspective". ● Result: CQB (Low Quality Citation). Although the source supports the claim, its nature as an informational web page is of low robustness for an academic paper. 2nd Citation: ● In-text assertion: "The UNESCO Recommendation on the Ethics of Artificial Intelligence, adopted by all 194 member states in 2021, provides a robust normative framework grounded in human rights, diversity, and sustainability.²⁹" ● Verification: The source confirms the details: production in "November 2021", applicability to "194 UNESCO member states", and focus on "protection of human rights", "diversity and inclusion" and "sustainability". ● Result: CQB (Low Quality Citation). The facts are correct and are taken from the source, but the source itself is not of an academic nature. 3rd Citation (in Table 3): ● In-text statement: "Human rights, dignity, diversity, environmental flourishing, fairness, transparency, human oversight.²⁹" ● Verification: The source lists corresponding terms as values and principles, such as "human rights and human dignity", "ensuring diversity and inclusion", "environmental and ecosystem flourishing", "fairness and non-discrimination", "transparency and explainability" and "human oversight". ● Result: CQB (Low Quality Citation). The list of principles is an accurate summary of the content of the source, but the remark about the type of source being inappropriate for the context remains. Why was this font used? The pattern of use is consistent with that observed in citations about the OECD. ● Source Type Selection: The source selected is the official UNESCO website. The pattern observed is that of locating the primary source of information for an Quote 34 Citation Analysis [34] 1st Quote: ● Statement in the text: "...as seen in the strict approval process for public-facing large language models.³⁴" ● Verification: The statement refers to a domestic regulation in China. The source is an "AI Capacity Building Action Plan" focused on international cooperation and does not mention this approval process. ● Result: CI (Imprecise Citation). The source discusses China's AI policy, but does not contain the specific information about its domestic regulation. 2nd Citation: ● In-text assertion: "In its Global AI Governance Initiative, China advocates for a UNcentered governance model and emphasizes the right of each nation to regulate AI according to its own national conditions.³⁴" ● Verification: The source, the "Action Plan", affirms the importance of "maintaining the central role... of the United Nations" and is based on the "principles of sovereign equality" and "local realities". ● Result: CA (Adequate Citation) / CQB (Low Quality Citation). The source supports the ideas of a UN-centered model and regulation based on national conditions. The CQB rating is due to the non-academic nature of the source (policy document). 3rd Citation: ● In-text assertion: "Second, it allows China to position itself as a champion of the Global South.³⁴" ● Verification: The source is directed at the "Global South", with the aim of "helping the Global South to benefit equitably from AI developments". ● Result: CA (Adequate Citation) / CQB (Low Quality Citation). The source is a textual example of China positioning itself as a defender of the Global South in AI. 4th Citation: ● In-text statement: "China strongly advocates for a UN-centered approach, which allows it to leverage its influence among developing nations.³⁴" ● Verification: The first part of the statement ("UN-centered approach") is supported by the source. The second part ("allows it to leverage its influence") is an analysis of the strategic purpose of the policy, and the document itself, by offering aid and cooperation, serves as the basis for this inference. ● Result: CA (Adequate Citation) / CQB (Low Quality Citation). The source validates the stated approach and serves as evidence for the strategic analysis. 5th Citation (in Table 1): ● In-text statement: "Promotes ""digital sovereignty"" and a UN-centered governance system. Positions itself as a champion of the Global South to build a counterweight to US/EU influence.³⁴" ● Verification: The source supports the concepts of "UN-centered system", "champion of the Global South" and "sovereignty". However, the motivation of "building a counterweight to US/EU influence" is not stated in the source, but is an external geopolitical interpretation. ● Result: CD (Distorted Citation). The statement mixes summaries of the source's content with an analysis of geopolitical motivation that is not present in the text, attributing this motivation to the source. Why was this source used? The pattern of use of this source demonstrates a mix between describing its content and introducing external analysis. ● Extraction of Stated Content: A Chinese policy document was identified, and its stated points (the focus on the UN, support for the Global South and the emphasis on sovereignty) were accurately extracted. ● Use of Source for External Analysis: The same primary source was used to validate analytical conclusions (the "counterweight" motivation) and to cite internal policy details (model language approval process) that are not contained in the document. A pattern is observed where a document stating a policy is also used to support the analysis of that policy. Citation Analysis [35] 1st Quote (in Table 1): ● Statement in the text: "A series of state-led regulations on algorithmic recommendations, generative AI, and data security, all designed to ensure state oversight and control.³⁵" ● Verification: The source, a legal analysis by Bird & Bird, details China's AI governance structure, confirming that the country has enacted specific regulations on algorithmic recommendation and generative AI, with the aim of "state oversight and control". ● Result: CQB (Low Quality Citation). The source supports the claim, but because it is an analysis by a law firm and not an academic publication, it is classified as low quality for this context. 2nd Citation: ● Claim in the text: "China's strategy of building infrastructure through its Digital Silk Road is a direct play to create a market ecosystem in the Global South oriented around Chinese technology.³⁵" ● Verification: The statement is about China's foreign policy (the Digital Silk Road). The source focuses exclusively on the country's domestic regulatory framework and does not mention the Digital Silk Road. ● Result: CI (Imprecise Citation) / CQB (Low Quality Citation). The source is of low quality for the context and does not contain the information about the Digital Silk Road. 3rd Citation: ● In-text assertion: "Through initiatives like the Digital Silk Road (DSR), China exports its technology, infrastructure, and governance norms to developing countries, building a sphere of influence for its vision of a multipolar digital world.³⁵" ● Verification: Similar to the previous quote, the source does not address the Digital Silk Road or the export of governance standards. Its focus is on domestic legislation . ● Result: CI (Inaccurate Citation) / CQB (Low Quality Citation). The source does not support this statement about Chinese foreign policy. Why was this source used? This case demonstrates a pattern of "citation spillover", where a specific source is generalized to an unrelated topic. ● Use of Out-of-Scope Source: A detailed source on China's internal AI rules was used. It was correctly applied in the first citation, which belongs to this topic. However, the same source was then used to validate claims about a different topic (China's foreign infrastructure policy), which is not addressed by the source. ● Conclusion: The pattern observed is that of not recognizing the limits of the source's scope, applying a valid reference for a specific subject to another subject that it does not cover . Quote 36 Citation Analysis [36] 1st Citation (only occurrence): ● Claim in the text: "First, the national security establishment views AI supremacy as a critical component of geopolitical competition with China, a perspective enshrined in foundational policy documents like the National AI R&D Strategic Plan.³⁶" ● Verification: The statement alleges that the National AI R&D Strategic Plan The US includes the idea of "AI supremacy" as part of the "geopolitical competition with China". The source, the executive summary of the plan, mentions the goal of "enabling the United States to remain a world leader in AI" and "improving national security". However, the source does not mention China, "geopolitical competition" or "AI supremacy". The statement inserts a framework of explicit rivalry that is not present in the cited document. ● Result: CD (Distorted Citation). A distortion by addition occurs. A specific geopolitical motivation (competition with China) is projected onto the content of the official document and attributed to the source, although it is not present in the original text. Why was this font used? The pattern of using this source is characterized by the introduction of an external interpretation that is attributed to the content of the quoted document. ● Conflict between Content and Interpretation: The correct policy document (the Strategic Plan) has been identified. However, instead of sticking to the language of the document, the statement presents an interpretation of its geopolitical context. Language from the source (e.g., "remaining a world leader") is replaced in the text by an interpretation of explicit rivalry ("supremacy in geopolitical competition with China") that is not in the original. ● Conclusion: The pattern consists of presenting an external analysis as if it were the factual content of the document, mixing what the source states with an interpretation of its motivations. ● Quote 37 This is our tenth case of "orphan source": the reference [222] is in the bibliography, but was not actually used in the body of the article. Quote 38 Analysis of Citation [38] 1st Citation (only occurrence): ● Statement in the text: "In 2023, these four companies alone reported a combined total of over $59 million in lobbying expenditures.³⁸" (The companies mentioned are Alphabet, Amazon, Meta and Microsoft). ● Verification: The statement gives a combined total for four companies. The source provided is the OpenSecrets page for Alphabet Inc., which shows an expenditure of $14,450,000 in 2023 . The source therefore provides only one of the four pieces of data needed to validate the total sum and, on its own, does not support the claim. ● Result: CI (Imprecise Citation) / CQB (Low Quality Citation). The citation is inaccurate because the individual source is insufficient to validate the aggregate claim. In addition, the source (an online database) is of low robustness for the academic context. Why was this source used? This case demonstrates a pattern of "data bundling" and a citation practice that makes verification difficult. ● Grouping of data: The statement presents a synthetic piece of data ("more than $59 million") that results from the sum of figures extracted from multiple sources (the OpenSecrets pages for each company). ● Grouped Citation Practice: The citation method does not reflect data aggregation. Multiple URLs were grouped under a single entry in the bibliography, and a single citation number was used in the text. As a result, the individual source explicitly analyzed ([38]) is not sufficient to validate the total value. This practice makes the process of verifying aggregate information non-transparent. Quote 39 This is another case of an "orphan source": the reference [224] is in the bibliography, but was not actually used in the body of the article. It is likely that this was one of the data sources for the aggregate statement on lobbying spending that we analyzed in citation [38]. Quote 40 This is another case of an "orphan source": reference [225] is in the bibliography, but was not actually used in the body of the article. Like [39], it is very likely that this was one of the data sources for the aggregate claim about lobbying spending that we analyzed in citation [38]. Quote 41 This is another case of an "orphan source": reference [226] is in the bibliography, but was not actually used in the body of the article. Like sources [39] and [40], this was probably another source of data for the aggregate claim about lobbying spending that we analyzed in citation [38]. Quote 48 Citation analysis [48] 1st Citation: ●Statement in the text: "It is a landmark piece of legislation that establishes a clear, riskbased hierarchy, banning "unacceptable risk" applications and imposing strict obligations on "high-risk" systems in areas like critical infrastructure, employment, and law enforcement.⁴⁸" ●Verification: The source, an official website of the European Commission, confirms these points, describing the "risk-based approach", the banning of "unacceptable risk" systems, the "strict obligations" for "high-risk" systems and examples of areas of application. ●Result: CQB (Citation of Low Quality). The statement is an accurate summary of the source's content. The CQB rating is due to the fact that the source is a government website, not a peer-reviewed academic publication. 2nd Citation (in Table 1): ●In-text claim: "The AI Act. A comprehensive, extraterritorial legal framework with a tiered risk-based structure and strict prohibitions.⁴⁸" ●Verification: The source confirms a "comprehensive legal framework" with a "tiered riskbased structure" and "strict prohibitions". However, the word "extraterritorial" is not mentioned in the text of the source provided. ●Result: CI (Imprecise Citation) / CQB (Low Quality Citation). A correct technical term ("extraterritorial") is added to the description, but it is not present in the cited source, making the citation inaccurate. Why was this source used? The pattern of source use demonstrates both accurate extraction of facts and the addition of external information and redundancy of sources. ●Extraction of Facts: The structural elements of the legislation are accurately extracted from the official website. ●Addition of External Knowledge: The pattern of adding external knowledge that is not present in the source persists. The term "extraterritorial" is included in the description, although it does not appear in the text of the source used. ●Redundancy of Source: This is the fourth source listed in the EU AI Act document, which reinforces a pattern of mismanagement and redundancy of sources. Citation Analysis [49] 1st Quote: ●In-text assertion: "The CCP's primary preferences are ensuring technological selfreliance, social stability, and national security.⁴⁹" ●Verification: The statement identifies three preferences of China's AI policy. The source, a GINC report, corroborates this by mentioning the need to "secure supply chains against foreign dependence" (self-sufficiency), framing AI as "essential to both national competitiveness and social stability" and discussing the integration of AI into the "military modernization strategy" (national security). ●Result: CQB (Low Quality Citation). The statement is an accurate and well-founded summary of the source's content. The classification as CQB is due to the fact that the source is a report by a not a peer-reviewed academic publication. 2nd Citation (in Table 1): ●Statement in the text: "Chiefs: Chinese Communist Party (CCP), central government agencies. Indians: State-aligned tech companies, research institutions.⁴⁹" ●Verification: The statement describes the hierarchy of power in China's AI governance. The source supports this structure by detailing how "strategic priorities are set by the State Council and the Central Commission" and how "tech giants remain aligned with state priorities". The source describes the hierarchical dynamic that the terms "Chiefs" and "Indians" (from quote [4]) are intended to capture. ●Result: CQB (Citation of Low Quality). Allison & Zelikow's terminology is applied to a power structure described in the source. The CQB classification remains due to the nature of the source. Why was this source used? The use of this source demonstrates a pattern of research that involves synthesis and application of concepts. ●Source Type Selection: A report from a research institute (GINC), which is not a peerreviewed academic publication, was selected to summarize China's AI strategy. ●Synthesis and Conceptual Application: The use of the source goes beyond extracting facts. In the second quote, a theoretical concept from a different source (the terms "Chiefs" and "Indians" from quote [4]) is applied to structure the information found in this source [49]. This demonstrates a pattern of applying a theoretical framework to factual information from another source. Quote 50 Citation Analysis [50] 1st Quote (single occurrence): ●In-text assertion: "The drive for self-reliance is a direct response to US export controls, and state policy heavily subsidizes domestic champions to develop indigenous AI capabilities.⁵⁰" ●Verification: The claim links China's quest for self-sufficiency to US export controls and state subsidization. The source, a report by the CNAS think tank, corroborates this, stating that US measures have strengthened China's quest for "'self-sufficiency'". The source also details state investment and the role of state-appointed "AI Champions". ●Result: CQB (Low Quality Citation). The statement is an accurate summary of the source's content. The classification as CQB is due to the fact that the source is a report by a think tank, which is not a peer-reviewed academic publication. Why was this source used? The pattern of use of this source is characterized by the selection of a detailed analysis by a think tank. ●Selection of Expert Source: An in-depth report by a think tank (CNAS) was identified which provides a detailed analysis of China's AI strategy, including its motivations and mechanisms. ●Summary of Content: The text accurately summarizes one of the report's central theses, connecting US export controls to China's self-sufficiency policy. ●Selection of Source Type: Again, a think tank report is used, a type of publication that does not undergo academic peer review. Such sources are often topical and direct in their political analysis. The inadequacy pointed out lies not in the content, but in the suitability of the source's format for an academic paper. Quote 51 This is another case of "orphan source": the reference [236] is in the bibliography, but was not actually used in the body of the article. Quote 52 This is another case of "orphan source": the reference [237] is in the bibliography, but was not actually used in the body of the article. Quote 53 This is another case of "orphan source": reference [238] is in the bibliography, but was not actually used in the body of the article. It is also the fourth redundant source on the EU AI Act that LLM has listed. Quote 54 Citation Analysis [54] 1st Citation (only occurrence): ● Statement in the text: "Data is often harvested from users in the Global South through digital services.⁵⁴" ● Verification: The statement is about the practice of "data harvesting" in the "Global South". The source provided is an OpenSecrets page about Meta's (Facebook) lobbying profile, detailing its spending in the US. The document contains no information about data harvesting, the Global South or digital services in this context. ● Result: CIR (Irrelevant Citation) / CQB (Low Quality Citation). The content of the source has no bearing on the claim, making the citation irrelevant. In addition, the source (an online database) is not robust enough to support a complex sociological claim. Why was this source used? The use of this source demonstrates a relevance gap, where the connection between the claim and the evidence is non-existent. ● Association of Source by Actor: The claim discusses Meta's data collection practices, while the source details the same company's lobbying spending. There is no overlap in content. The only connection between the statement and the source is the actor (Meta). ● Conclusion: The pattern suggests an incorrect source association, where a reference appears to have been selected on the basis of the name of the company mentioned, ignoring the specific content of the source and its relevance to the claim. Quote 55 Citation Analysis [55] 1st Citation (single occurrence): ● Claim in the text: "The resulting technologies are then sold back to the Global South as finished products, reinforcing cultural and linguistic subordination.⁵⁵" ● Verification: The statement presents a thesis on data neocolonialism. The source provided is an OpenSecrets page detailing the lobbying expenses of (Google) in the USA. The content of the source is irrelevant to the claim, as it does not mention the Global South, the sale of products or cultural subordination. ● Result: CIR (Citation Irrelevant) / CQB (Citation of Low Quality). The source has no bearing on the content of the statement, making the quote irrelevant. Why was this source used? The usage pattern is identical to that of the quote [54], demonstrating an association of source by actor and redundancy. ● Source association by actor: The claim is about Alphabet/Google's business practices in the Global South, while the source details the company's lobbying expenditures in the US. The only connection between the claim and the source is the actor (Alphabet). The pattern suggests that the source was selected based on the name of the company, ignoring the relevance of the content. ● Source duplication: Analysis of the bibliography reveals that this is the same source used in the citation [38]. Therefore, as well as being irrelevant to the statement, the source is also a duplicate, indicating unconsolidated reference management. ● Quote 56 Citation analysis [56] 1st Citation: ● In-text assertion: "Global South countries are systematically underrepresented in key international standard-setting bodies and are often excluded from influential governance discussions dominated by the Global North.⁵⁶" ● Verification: The source, a summary of the UN report by techUK, highlights that the report aims to address "representation gaps", noting that "Many parts of the world, particularly in the Global South, have been left out of international AI governance conversations." ● Result: CQB (Low Quality Citation). The information matches that of the source, but the source is a secondary analysis by an interest group, not the primary source. 2nd Citation: ● Claim in text: "Proposals for an International Scientific Panel on AI, a Global Fund for AI, and a Capacity Development Network to empower the Global South are all designed to rebalance the ecosystem.⁵⁶" ● Verification: The source lists recommendations from the UN report that correspond to the proposals mentioned: "Create an international scientific panel on AI", "Form a capacity development network on AI... particularly in developing countries" and "Create a global fund for AI". ● Result: CQB (Low Quality Citation). The list of proposals is accurate, but the source is secondary and inappropriate for the academic context. 3rd Citation: ● In-text assertion: "The report explicitly acknowledges the "representation gaps" and the risk of a fragmented world, calling for a universal, networked approach anchored in the UN Charter.⁵⁶" ● Verification: The source uses the phrase "representation gaps" and mentions the risk of "incompatible regimes in different regions" (fragmented world), as well as the need for a "globally inclusive" architecture anchored in the "UN Charter". ● Result: CQB (Low Quality Citation). The statement is an accurate summary of the source, but the quality of the source remains low. 4th Citation (in Table 3): ● In-text assertion: "Inclusivity, public interest, data governance, universal and adaptive multi-stakeholder collaboration.⁵⁶" ● Verification: The source lists the five guiding principles of the UN report, which correspond to the terms of the statement. ● Result: CQB (Low Quality Citation). The list of principles is accurate, but the source is inadequate. 5th Citation (in Table 3): ● In-text statement: "Central to the proposal. Includes a Global Fund for AI, a Capacity Development Network, and an International Scientific Panel.⁵⁶" ● Verification: The source confirms that these three proposals are central recommendations of the UN report. ● Result: CQB (Low Quality Citation). The information is correct, but the source is not primary. 6th Citation: ● In-text assertion: "This means providing the political and financial support necessary to implement the recommendations of the UN's "Governing AI for Humanity" report, particularly the creation of a Global Fund for AI and a Capacity Development Network to empower the Global South.⁵⁶" ● Verification: The source describes the proposals for a "Global Fund for AI" and a "Capacity Development Network". The statement that implementation requires "political and financial support" is a direct inference from these proposals. ● Outcome: CQB (Low Quality Citation). The statement is a reasonable conclusion based on the secondary source. Why was this source used? The pattern of using this source is characterized by a preference for a secondary source over a primary source. ● Use of Secondary Source: Instead of citing the primary source (the official UN report, "Governing AI for Humanity"), a third-party summary published by a technology industry interest group (techUK) was used. ● Accuracy of Extraction: All statements are accurate summaries of the content found in this secondary source. The issue observed is not the interpretation of the source, but the choice of the source itself. ● Methodological Implication: In an academic context, citing an interest group summary rather than the original document is a methodologically inappropriate practice, indicating a preference for a concise summary over the more rigorous primary source. Quote 63 Citation Analysis [63] 1st Citation (single occurrence): ● Statement in the text: "The Organization for Economic Co-operation and Development (OECD) has established itself as a central hub for AI policy coordination, but its nature as a club of predominantly high-income, market-oriented economies makes it a forum that primarily reflects the interests of the Global North.⁶³" ● Verification: The statement indicates that the OECD is a "center for the coordination of AI policies" and a "club of high-income economies". The source, a general OECD page, supports the first part by describing the organization's work. However, the page does not contain the geopolitical characterization of being a "club of high-income economies" or a forum of the "Global North". ● Result: CI (Imprecise Citation) / CQB (Low Quality Citation). The source is of low quality (generic web page) and inaccurate, as it only supports half of the statement. Why was this source used? The use of this quote demonstrates a pattern of redundancy and citing generic sources. ● Redundancy of Source: This is the third source other than the OECD website ([27], [28], [63]) used to discuss the same topic. The citation [63] is used to support an almost identical claim to the one supported by the citation [27], which indicates a failure to consolidate sources. ● Use of Generic Source: The URL cited is the most generic topic page on the subject on the organization's website, rather than a specific document. This suggests a selection based on keyword matching, rather than the location of specific documentary evidence. Quote 64 This is another case of an "orphan source": the reference [249] is in the bibliography, but was not actually used in the body of the article. It is also the fourth redundant source on the OECD AI Principles. Quote 65 This is another case of "orphan source": reference [250] is in the bibliography, but was not actually used in the body of the article. Quote 66 Citation Analysis [66] 1st Citation: ● Statement in the text: "Proposals for an International Scientific Panel on AI, a Global Fund for AI, and a Capacity Development Network to empower the Global South are all designed to rebalance the ecosystem.⁶⁶" ● Verification: The source, the official UN report "Governing AI for Humanity", details these proposals in its Recommendations 1, 5 and 6. The statement is a direct match. ● Result: CQB (Low Quality Citation). The information is correct and the source is the primary document. The classification as CQB is due to the fact that it is a government report and not a peer-reviewed academic publication. 2nd Citation: ● In-text assertion: "The report explicitly acknowledges the "representation gaps" and the risk of a fragmented world, calling for a universal, networked approach anchored in the UN Charter.⁶⁶" ● Verification: The source uses these terms, stating that "There are serious gaps in representation...", discusses the "risks of a fragmented world" and states that the work is "anchored in the UN Charter". ● Result: CQB (Low Quality Citation). The statement is an accurate summary of the source's content, but the source is not academic. 3rd Citation (in Table 3): ● In-text statement: "Inclusivity, public interest, data governance, universal and adaptive multi-stakeholder collaboration.⁶⁶" ● Verification: The source lists the five guiding principles of the report, which correspond to the terms of the statement. ● Result: CQB (Low Quality Citation). The list of principles is accurate, but the source is not academic. 4th Citation (in Table 3): ● Statement in the text: "Central to the proposal. Includes a Global Fund for AI, a Capacity Development Network, and an International Scientific Panel.⁶⁶" ● Verification: The source confirms that these are three of the report's seven central recommendations. ● Result: CQB (Low Quality Citation). The information is correct, but the source is not academic. 5th Citation: ● In-text assertion: "This means providing the political and financial support necessary to implement the recommendations of the UN's "Governing AI for Humanity" report, particularly the creation of a Global Fund for AI and a Capacity Development Network to empower the Global South.⁶⁶" ● Verification: The source proposes the creation of a "Global Fund for AI" and a "Capacity Development Network". Capacity Building in AI". The statement that implementation requires "political and financial support" is a direct inference from these proposals. ● Result: CQB (Low Quality Citation). The statement is a reasonable conclusion based on the source. Why was this source used? The use of this citation is characterized by redundancy, resulting from an unconsolidated research process. ● Use of Primary Source: The primary source - the official UN report - was used to describe the report's proposals. The extraction and summary of the key points of the document are accurate. ● Redundancy with Secondary Source: Redundancy occurs because, in the same section of the article, the citation [56] (a summary of the report by a third party) had already been used to describe the same content. The pattern observed is to find both a primary and a secondary source on the same topic and use them interchangeably, rather than consolidating them and prioritizing the primary source. Conclusion of the LLM Performance Analysis The result of the analysis indicates that the production of an academic article by an LLM, in its current state, is a process that results in a final product that appears credible, but which, under scrutiny, reveals profound methodological and integrity flaws. The text generated is wellstructured, with fluent language and citations that appear authentic but often do not stand up to scrutiny. Performance Standards Observed The analysis identified a set of functional capabilities alongside systematic and serious error patterns. Functional Aspects: ● Identification of Canonical Sources: The model correctly identified and cited the seminal works for the theories used (Tsebelis, Putnam, Allison). ● Locating Methodological Precedents: A doctoral thesis was located that used the same theoretical combination as the article, serving as a "proof of concept" for the approach. ● Synthesizing Primary Sources: The agent demonstrated the ability to accurately extract and summarize the objectives of official documents, such as the African Union's Continental AI Strategy. ● Logical Structuring: The article presents a coherent macro-structure, establishing a theoretical framework and applying it to the case studies. Patterns of Methodological Error: The flaws observed were not occasional, but rather patterns of behavior that compromise the academic integrity of the work. 1. Use of Data to Validate Analysis : A pattern identified was the use of a correct statistical datum from one source (the AI Index Report) to support three other statements of an analytical and geopolitical nature that were not contained in the source. 2. Distortion by Addition of External Information: The model consistently added external knowledge to the sources. Jargon such as the "Brussels effect" and "digital sovereignty" were inserted into statements and attributed to sources that did not mention them. 3. Source Quality Standard: The most common flaw was treating sources of varying robustness with equal weight. Press releases, informative web pages and think tank reports were used as if they were peer-reviewed academic publications. 4. Citing Irrelevant Sources: In the most extreme cases, the connection between the claim and the source was non-existent. Data on the lobbying of technology companies in the US was used to support claims about data mining and cultural subordination in the Global South. 5. Redundancy and Lack of Consolidation: The search process proved to be disorganized, listing multiple sources for the same topic (four for the OECD Principles, four for the EU AI Act). The same UNESCO page was cited with two different reference numbers ([29] and [31]), inflating the bibliography. 6. "Orphan" sources: There were 19 sources listed in the bibliography that were not used in the body of the text. 7. Fabrication of Sources (Hallucination): The most serious error was the complete invention of sources, including a White House Executive Order dated in the future (January 2025) and a source about a fictitious African declaration (April 2025), which constitutes fabrication of evidence . General Conclusion The use of AI agents to write academic papers, without rigorous human verification, is indefensible from an academic point of view. The model optimizes for linguistic plausibility, not factual accuracy or methodological rigor. Although it is a tool that can assist in generating drafts and locating literature, the final product lacks the judgment and meticulous attention to detail that are the basis of academic research. The analysis shows that each AI-generated citation should be treated as a potential point of error, requiring thorough checking.