Hennighausen, Ch is ine; Ya za Na a o‐Schä , Vanessa Gab iela; Elle , E ic
A icle — Published Ve sion
AI‐Media ed Communica ion in E‐Comme ce: Implica ions
o Cus ome T us
In e na ional Jou nal o Consume S udies
P o ided in Coope a ion wi h:
John Wiley & Sons
Sugges ed Ci a ion: Hennighausen, Ch is ine; Ya za Na a o‐Schä , Vanessa Gab iela; Elle , E ic
(2025) : AI‐Media ed Communica ion in E‐Comme ce: Implica ions o Cus ome T us , In e na ional
Jou nal o Consume S udies, ISSN 1470-6431, Wiley, Hoboken, NJ, Vol. 49, Iss. 5,
h ps://doi.o g/10.1111/ijcs.70111
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/330201
S anda d-Nu zungsbedingungen:
Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen
Zwecken und zum P i a geb auch gespeiche und kopie we den.
Sie dü en die Dokumen e nich ü ö en liche ode komme zielle
Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich
machen, e eiben ode ande wei ig nu zen.
So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen
(insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en,
gel en abweichend on diesen Nu zungsbedingungen die in de do
genann en Lizenz gewäh en Nu zungs ech e.
Te ms o use:
Documen s in EconS o may be sa ed and copied o you pe sonal
and schola ly pu poses.
You a e no o copy documen s o public o comme cial pu poses, o
exhibi he documen s publicly, o make hem publicly a ailable on he
in e ne , o o dis ibu e o o he wise use he documen s in public.
I he documen s ha e been made a ailable unde an Open Con en
Licence (especially C ea i e Commons Licences), you may exe cise
u he usage igh s as speci ied in he indica ed licence.
h ps://c ea i ecommons.o g/licenses/by/4.0/
1 o 17
In e na ional Jou nal o Consume S udies, 2025; 49:e70111
h ps://doi.o g/10.1111/ijcs.70111
In e na ional Jou nal o Consume S udies
ORIGINAL ARTICLE OPEN ACCESS
AI- Media ed Communica ion in E- Comme ce: Implica ions
o Cus ome T us
Ch is ineHennighausen1 | VanessaGab ielaYa zaNa a o-Schä 2 | E icElle 1
1Business School, Technische Hochschule Ingols ad , Ingols ad , Ge many | 2Cen e o Leade ship and People Managemen , Ludwig- Maximilian-
Uni e si ä München, München,Ge many
Co espondence: Ch is ine Hennighausen (ch is ine.hennighausen@ hi.de)
Recei ed: 28 May 2024 | Re ised: 15 May 2025 | Accep ed: 4 Augus 2025
Funding: The open access publica ion o his wo k was suppo ed by he Open Access Publica ion Fund o Technische Hochschule Ingols ad (THI).
Keywo ds: AI- media ed communica ion (AI- MC)| a i icial in elligence| CRM| cus ome us | E- comme ce| gene a i e AI| se ice c i icali y
ABSTRACT
Gene a i e a i icial in elligence (AI) echnologies o e new po en ial o ma ke ing and cus ome ope a ions, such as au o-
ma ion and pe sonaliza ion o cus ome se ice. Howe e , mo e mus be unde s ood abou how AI- media ed communica ion
(AI- MC) a ec s cus ome us . We conduc ed an online expe imen o in es iga e he impac o AI- MC on cus ome us in an
online e ail con ex . We p esen ed N = 294 pa icipan s wi h wo email scena ios desc ibing a p oduc e u n con ex , labeled
as w i en by ei he (a) a se ice employee, (b) a se ice employee assis ed by AI, o (c) AI on behal o he se ice employee.
We u he a ied le els o se ice c i icali y o conside cus ome s' pe cep ion o ulne abili y. Ou indings e ealed highe
cus ome us a ings in he online e aile when he email communica ions we e w i en by he se ice employee, compa ed
o hose w i en by he se ice employee assis ed by AI. When analyzing he di e en componen s o us , i was ound ha
communica ions w i en by he se ice employee assis ed by AI educed pe cep ions o bo h he online e aile 's bene olence and
in eg i y, while communica ions w i en by AI on behal o he employee led o lowe pe cei ed in eg i y o he online e aile .
Su p isingly, se ice c i icali y did no a ec us a ings. We discuss he manage ial implica ions o in eg a ing gene a i e AI
in o cus ome se ice in he con ex o he EU AI Ac , which came in o o ce on 1 Augus 2024.
1 | In oduc ion
Wi h he elease o Cha GPT- 3 in No embe 2022 (C aw o d
e al. 2023), a new e a has begun. A i icial in elligence (AI)
cha bo s can now gene a e answe s ha a e language- wise
co ec , compelling, and ha d o dis inguish om human-
gene a ed con en (Mei e al.2024). Due o i s ease o use and
po en ial o inc ease e iciency, Cha GPT has been widely
adop ed in bo h p i a e (Tho mudsson 2023a) and business
con ex s (Tho mudsson2023b). In ma ke ing and cus ome e-
la ionship managemen (CRM; i.e., in e ac ions and p ocesses
acili a ed by a business o manage and nu u e ela ionships
wi h cus ome s h ough he in eg a ion o people, p ocesses,
and echnology; Chen and Popo ich2003), he eno mous busi-
ness po en ial o gene a i e AI (i.e., AI sys ems ha can au ono-
mously gene a e con en , such as ex o images, using ad anced
machine lea ning models ained on la ge da ase s; Kaplan
and Haenlein2019) is e lec ed in excep ionally high adop ion
a es (Denche a 2023; Tho mudsson 2023b), o ins ance, o
au oma izing cus ome communica ion in social media and
email messaging (Schweidel e al.2023). Unde he EU AI Ac ,
which came in o o ce on 1 Augus 2024, Eu opean compa-
nies a e obliged o label AI- gene a ed con en and make i s use
anspa en o hei cus ome s (Eu opean Commission2024).
This is an open access a icle unde he e ms o he C ea i e Commons A ibu ion License, which pe mi s use, dis ibu ion and ep oduc ion in any medium,
p o ided he o iginal wo k is p ope ly ci ed.
© 2025 The Au ho (s). In e na ional Jou nal o Consume S udies published by John Wiley & Sons L d.
B ie abs ac s o hese indings ha e been submi ed o p esen a ion a he 53 d Cong ess o he Ge man Psychological Socie y (DGPs)/15 h Con e ence o he
Aus ian Psychological Socie y (ÖGP) 2024, 16- 19 Sep embe and a he 22nd Cong ess o Wo k and O ganiza ional Psychology (EAWOP) 2025, 21- 24 May.
2 o 17 In e na ional Jou nal o Consume S udies, 2025
Such egula ions e lec a b oade in e na ional discou se on AI
go e nance, highligh ing he need o adap legal amewo ks,
including da a secu i y and p o ec ion, as well as s eng hen
e hical ounda ions in esponse o he challenges posed by AI
(Al Najdawi e al.2024, 2025). This aises he ques ion o how
cus ome s eac owa d he explici use o AI and AI- media ed
communica ion (AI- MC) in ma ke ing and CRM and whe he
anspa ency abou AI usage enhances o unde mines cus ome
us (Bakonyi2024; Sigala e al.2024).
P e ious s udies showed ha cus ome s a e o en skep ical
when in e ac ing wi h AI. Fo ins ance, cus ome pu chase a es
d opped by nea ly 80% when he iden i y o an AI cha bo was
e ealed a he beginning o a sales call compa ed o when he
iden i y o he AI cha bo was no disclosed (Luo e al.2019). A
ecen s udy by Liu e al.(2022) showed ha indi iduals mis-
us he sende o an email dealing wi h a p oduc inqui y when
he email has been w i en wi h he help o AI. Resea ch on he
“Wo d- o - Machine” e ec sugges s ha cus ome s' p e e ences
o AI o human ecommenda ions depend on whe he he de-
cision con ex is u ili a ian o hedonic. Cus ome s pe cei e AI
as mo e compe en han humans o u ili a ian asks bu less so
o hedonic asks (Longoni and Cian2020). This highligh s ha
AI- d i en ecommenda ions in u ili a ian se ings may os e
us , while hey could e ode us in hedonic scena ios. Fu he
e idence sugges s ha cus ome s end o mis us AI, especially
insi ua ions o high se ice c i icali y, ha is, when he ou come
o a se ice in e ac ion is signi ican o he cus ome (Chen e al.
2022; Xu, Shieh, e al.2020). High se ice c i icali y o en leads
o inc eased cus ome ulne abili y, as he ou comes o se ice
in e ac ions can signi ican ly impac cus ome s' well- being o
sa is ac ion (Dehghanpou i e al.2020). Moza a i e al.(2021)
ound ha i se ice c i icali y is high, disclosing a cha bo 's
iden i y as such nega i ely a ec ed cus ome us in he con-
e sa ional pa ne , while cus ome us was no impai ed
when se ice c i icali y was low. This aligns wi h indings om
Logg e al.(2019), who disco e ed ha despi e ini ial conce ns
abou algo i hmic decisions, indi iduals o en exhibi algo i hm
app ecia ion, meaning hey us algo i hmic judgmen s mo e
han human ones in ce ain domains, especially when accu acy
is emphasized. Since he widesp ead use o gene a i e AI is qui e
a new phenomenon, howe e , esea ch has jus begun o in es-
iga e cus ome s' eac ions owa d AI and AI- MC in ma ke ing
and CRM.
To make a signi ican and no el con ibu ion o he li e a u e
on AI- MC and cus ome us , ou s udy aims o del e deepe
in o he unde s anding o cus ome s' pe cep ions o AI- MC in an
e- comme ce si ua ion, speci ically ocusing on cus ome us ,
using a ealis ic CRM scena io. We, he e o e, conduc ed an ex-
pe imen al s udy a ying he sende o an email in a p oduc
e u n con ex om no AI agency o comple e AI agency. We
desc ibed he email communica ion wi h he cus ome as ei he
w i en by (a) a se ice employee, (b) a se ice employee assis ed
by AI, o (c) an AI on behal o he se ice employee. We u he
modi ied he se ice c i icali y o he p oduc e u n scena io by
desc ibing ei he a high- p ice p oduc u gen ly needed o a low-
p ice p oduc no u gen ly demanded. As highligh ed by ecen
AI esea ch, managing AI- d i en communica ions in a way ha
main ains consume au onomy and us is c i ical o a oiding
nega i e eac ions (Spais e al. 2023). By examining he us
componen s bene olence, in eg i y, and compe ence, we gain
a deepe unde s anding o how AI- MC a ec s cus ome us
and hus make bo h heo e ical con ibu ions (i.e., by ex ending
he ecen li e a u e on how di e en deg ees o AI- MC impac s
cus ome us , speci ically by disen angling us in i s compo-
nen s; Akba e al.2024) and p ac ical con ibu ions (e.g., o
de eloping ma ke ing s a egies o inco po a e AI- MC in CRM
e ec i ely). Ou esea ch hus p o ides a no el and signi ican
con ibu ion o he cu en li e a u e on AI- MC and cus ome
us , which is o high con empo a y ele ance, pa icula ly in
ligh o cu en egula ions emphasizing anspa ency in he use
o AI in ma ke ing (Eu opean Commission2024).
2 | Theo e ical Backg ound
Wi h he new possibili ies o gene a i e AI c ea ing ex s, pic-
u es, and audio isual ma e ial, he ques ion a ises abou how
AI- MC a ec s in e pe sonal communica ion and migh change
he way con e sa ional pa ne s pe cei e each o he . In e-
comme ce, elici ing and p o ec ing us , i.e., “ he willingness o
a pa y o be ulne able o he ac ions o ano he pa y based on
he expec a ion ha he o he will pe o m a pa icula ac ion
impo an o he us o , i espec i e o he abili y o moni o
o con ol ha o he pa y” (Maye e al.1995; 712), is a p inci-
pal goal, as cus ome us is essen ial o a company's success
(Isae a e al.2020). Howe e , he apid p oli e a ion o gene a-
i e AI along he cus ome jou ney (Dencik e al.2023) is aced
wi h he ecen obse a ion ha cus ome s o en mis us AI-
gene a ed communica ion (P akash e al.2023). Cus ome s may
ind hemsel es in a ulne able posi ion while in e ac ing wi h
AI due o hei pe sonal ci cums ances o he complexi ies o a
pa icula si ua ion (Mogaji e al.2020). The pe cep ion and use
o AI could po en ially exace ba e his ulne abili y. As no ed by
Longoni and Cian(2020) AI- MC is mo e eadily us ed when
cus ome s iew he con ex as u ili a ian, whe e AI's compe-
ence shines, a he han hedonic, whe e cus ome s may doub
AI's emo ional in elligence and nuanced decision- making.
While AI is me ely a ool wi h no inhe en ha m ul a ibu es
(Tu ley2019), i s usage in cus ome se ice is o en segmen ed
based on ask complexi y. Cus ome s seem o p e e AI o less
challenging issues bu lean owa d human assis ance when asks
become mo e in ica e (Xu, Chen, e al.2020). The p e e ence is
d i en by cus ome s' belie in he AI's capaci y o sol e p oblems
(P akash e al.2023). Howe e , his can po en ially ampli y he
cus ome 's ulne abili y i AI ails o sol e complex issues, esca-
la ing he pe cei ed isk (Xu, Chen, e al.2020). Fu he mo e,
AI's g owing ole in ma ke ing has aised conce ns abou i s
po en ial o unde mine cus ome au onomy. The abili y o au-
oma e decisions and ecommenda ions wi hou human o e -
sigh can some imes lead o pe cep ions o con ol loss, u he
e oding us in AI (Spais e al.2023). In heal hca e con ex s, o
example, cus ome s a e o en esis an o AI ecommenda ions
due o pe cei ed uniqueness neglec — he belie ha AI sys ems
canno accoun o hei indi idual ci cums ances (Longoni
e al.2019). These indings ein o ce he impo ance o us in
AI- MC in scena ios whe e he pe cep ion o pe sonal unique-
ness is c ucial. AI- MC can disappoin cus ome s' expec a ions
o companies and e oke ea s o being decei ed o manipula ed
(Jakesch e al.2019). In line wi h his, cus ome s can pe cei e
ex s w i en wi h he aid o AI o by AI as less us wo hy
3 o 17
(Hancock e al.2020). Taking his a s ep u he , AI's pe cei ed
con ol o e he cus ome , o lack he eo , can also in luence
a cus ome 's us in AI agen s (Yang e al.2022). The au ho s
disco e ed ha , while an h opomo phizing AI agen s can en-
hance cus ome s' accep ance in cases whe e he pe cei ed con-
ol is high—i could con e sely dec ease hei accep ance in
low con ol si ua ions—adding ye ano he sou ce o ulne a-
bili y. Jakesch e al.(2019) in es iga ed how using AI o c ea e
online p o iles o Ai bnb hos s in luenced us pe cep ions o
hese hos s. Thei indings sugges ha when p o ile desc ip-
ions o he Ai bnb hos s we e p esen ed as ei he all human-
w i en o all AI- gene a ed, pa icipan s ga e almos iden ical
us a ings o he p o iles ma ching he p e- a ed us le el
o he p o iles. In a se ing whe e pa icipan s suspec ed hos
p o iles o be AI- gene a ed o labeled as such, howe e , pa ic-
ipan s a ed hese as less us wo hy. The au ho s e e o his
obse a ion as Replican E ec and explain hei indings wi h
he Hype pe sonal Model o CMC (Wal he and Whi y2021),
p oposing ha o e - a ibu ions based on he pa icipan s' suspi-
cion ha he p o ile was AI- gene a ed o by he label indica ing
ha a p o ile was AI- gene a ed accoun s o he lowe us a -
ings. Liu e al.(2022) explo ed AI- MC in email communica ions,
a ying he deg ee o AI- MC and he in e pe sonal emphasis o
he email communica ion. They ound ha pa icipan s a ed
hei con e sa ional pa ne lowes on us wo hiness when
AI comple ely media ed he email communica ion, while us
pe cep ions we e highe when he con e sa ional pa ne was
aided by AI o used no AI in his e- mail message. In e es ingly,
his e ec occu ed independen o he pe sonal emphasis o he
communica ion. D awing on he desc ibed indings, we expec
simila esul s o an e- comme ce con ex and hypo hesize:
H1. Cus ome us in an e- com e aile is highe when an on-
line e aile 's email is labeled as being w i en by a se ice em-
ployee han when he email is labeled as being w i en by a se ice
employee assis ed by AI o by AI on behal o he se ice employee.
T us is a mul i- dimensional cons uc comp ising a leas h ee
elemen s: bene olence, in eg i y, and compe ence (McKnigh
and Che any2001). Bene olence e e s o he deg ee o pos-
i i e o ien a ion o a us wo hy pa y owa d he wel a e o
he us beyond sel ish gain (Maye e al.1995; McKnigh and
Che any 2001). In eg i y e eals i sel in he pe cep ion ha
he us wo hy pa y adhe es o ecognized p inciples and
alues ha align wi h he us o 's expec a ions (McKnigh
and Che any 2001; Moza a i e al. 2021). Finally, compe-
ence ela es o he abili y o he us wo hy pa y o pe o m
ce ain ac i i ies sa is ac o ily (Moza a i e al.2021; Ro iq and
Mula2009). Recen esea ch elucida es ha AI sys ems o en ap-
pea highly compe en bu less bene olen and less in eg al han
hei human coun e pa s, which poses an in luence on us o -
ma ion (Li and Bi e ly2024; No ozhilo a e al.2024). In some
con ex s, AI sys ems a e e en pe cei ed as mo e compe en han
humans, which means ha people a e mo e likely o ollow he
ecommenda ions o AI han o o he people (Spais e al.2023;
Zhang e al.2023). The le el o agency acco ded o AI can bo h
enhance and diminish us (Vannes e and Pu anam 2021).
This p omp s he need o a p ecise conside a ion o us com-
ponen s when adop ing AI sys ems in se e al socie al se ings
(No ozhilo a e al. 2024). Thus, acknowledging and unde -
s anding cus ome ulne abili y in he blend o human and AI
cus ome se ice in e ac ion is c ucial o ensu e he success-
ul managemen o consume us in he ealm o AI- MC and
consequen ly in he success o e- comme ce businesses (Ameen
e al.2021). Moza a i e al.(2021) ound ha he loss o us in
cha bo s a e disclosu e insi ua ions o high se ice c i icali y
was due o lowe pe cep ions o he cha bo 's compe ence and
bene olence bu no based on lowe pe cep ions o he cha bo 's
in eg i y. Mo eo e , when a cus ome 's se ice eques could no
be esol ed, cha bo disclosu e mi iga ed he ad e se cus ome
eac ions so ha cus ome us was e en inc eased (Moza a i
e al.2021). In pa icula , cus ome s eac ed posi i ely o cha -
bo disclosu e in a ailu e si ua ion due o i me belie s in he
cha bo 's in eg i y and bene olence, while compe ence belie s
we e no a ec ed. AI- powe ed se ices in shopping ha e been
shown o imp o e he o e all cus ome expe ience by inc easing
us , con enience, pe sonaliza ion, and ela ionship engage-
men while educing pe cei ed sac i ices (Ameen e al.2021).
Howe e , his also depends on how compe en ly he AI sys-
em pe o ms i s asks—AI cus ome se ice has been ound o
be a o ed o asks wi h low se ice c i icali y, while human
cus ome se ice is a o ed o asks wi h high se ice c i ical-
i y, sugges ing ha pe cei ed p oblem- sol ing abili y plays an
impo an ole in media ing cus ome s' usage in en ions (Xu,
Shieh, e al.2020). Heide 's(1958) a ibu ion heo y posi s ha
indi iduals make sense o hei su oundings by a ibu ing
cause and e ec o people's beha io s and e en s. In doing so,
indi iduals a e emp ed o ela e he beha io o o he s o in e -
nal ac o s such as pe sonali y ai s o abili ies (Heide 1958).
Recen esea ch showed ha con e sa ional pa ne s using AI
in in e pe sonal ela ionships we e no conside ed o pu he
same e o in o a iendship as hey would by pe sonally w i -
ing he message, which, in u n, leads o dec eased ela ionship
sa is ac ion and unce ain y o he con e sa ional pa ne (B.
Liu e al.2023). Acco dingly, we a gue ha in CRM con ex s,
cus ome s may pe cei e online e aile s as less bene olen and
in eg al i a se ice employee, ep esen ing he e aile , pa ly
o comple ely uses AI o engage wi h he cus ome , as he use
o AI may imply diminished e o om he se ice employee
in communica ion compa ed o manual d a ing (P en ice and
Nguyen2020). We hus hypo hesize:
H2. Cus ome s' pe cep ions o bene olence and in eg i y a e
highe when an online e aile 's email is labeled as being w i en
by a se ice employee han when he email is labeled as being w i -
en by a se ice employee assis ed by AI o by AI on behal o he
se ice employee.
Since us enables indi iduals o accep pe cei ed ulne abil-
i y (Maye e al.1995), cus ome us seems o be especially
ele an in cus ome si ua ions o high se ice c i icali y, ha
is, when he ou come o a se ice si ua ion pa icula ly ma e s
o cus ome s (C isa ulli and Singh2017). Moza a i e al.(2021)
showed ha in an AI- Cha bo en i onmen , se ice c i icali y
mode a ed he us –AI ela ionship ac oss wo sepa a e ex-
pe imen s. Insi ua ions o high se ice c i icali y, pa icipan s
us ed less in a con e sa ional pa ne when his pa ne dis-
closed he sel as a cha bo a he end o he con e sa ion com-
pa ed o a non- disclosu e scena io. Howe e , cha bo disclosu e
did no a ec cus ome a ings in a si ua ion o low se ice c i -
icali y bu in a si ua ion o high se ice c i icali y. In con as ,
when he cha bo could no sol e he cus ome 's se ice eques ,
4 o 17 In e na ional Jou nal o Consume S udies, 2025
cha bo disclosu e enhanced cus ome us . I mi iga ed he e -
ec s o he nega i e se ice ou come as compa ed o when he
cha bo was no disclosed. In line wi h hese indings, we expec
se ice c i icali y o mode a e he ela ionship be ween AI con-
di ion and cus ome us , especially insi ua ions whe e se ice
employee communica ion is pa ially o ully media ed by AI.
Mo e speci ically, we sugges ha insi ua ions o high se ice
c i icali y, he di e en amoun s o AI in a communica ion will
lead o la ge di e ences in cus ome us han hose o low se -
ice c i icali y. We hus hypo hesize:
H3. Se ice c i icali y mode a es he ela ionship be ween
AI- MC and cus ome us in an online e aile , such ha he
di e ence in cus ome us be ween AI- MC condi ions is mo e
p onounced when se ice c i icali y is high compa ed o when se -
ice c i icali y is low.
3 | Ma e ials & Me hods
3.1 | Design and P ocedu e
We conduc ed an online expe imen o in es iga e how di e en
deg ees o AI- MC a ec ed cus ome us in an online e aile
and how much his ela ionship di e s be ween high and low
se ice c i icali y scena ios. The expe imen ollowed a 3 (deg ee
o AI- MC: email communica ion w i en by se ice employee s.
w i en by se ice employee assis ed by AI s. w i en by AI on
behal o he se ice employee) ×2 (se ice c i icali y: high s.
low) be ween- subjec s design. To educe demand e ec s and
o conceal he pu pose o ou s udy, he consen o m b ie ed
pa icipan s ha he s udy aimed o explo e hei pe cep ions
and eac ions o email communica ion om an online e aile .
Pa icipan s we e assigned andomly o an expe imen al condi-
ion and p esen ed wi h an in oduc o y ex and a sc eensho o
an email exchange wi h an online e aile . They we e ins uc ed
o pu hemsel es in he cus ome 's shoes, isualizing he ou -
lined scena io. Nex , pa icipan s we e asked o e alua e hei
us in he online e aile . T us a ings se ed as dependen
a iables. Pa icipan s we e hen asked o indica e hei pe cep-
ions o se ice c i icali y and ealism o he scena io as a manip-
ula ion and alidi y check, espec i ely. On he ollowing pages
o he ques ionnai e, da a on he pa icipan s' disposi ional us
and a i ude owa d AI we e collec ed as con ol a iables, as
well as demog aphic in o ma ion (gende , age, highes le el o
educa ion, cu en oca ional s a us, ield o s udy/p o ession).
Be o e he end o he ques ionnai e, pa icipan s we e asked
o indica e how he email communica ion p esen ed had been
w i en as an a en ion check. The a en ion check addi ionally
ensu ed ha he esponden s unde s ood whe he he email
was w i en by a se ice employee, an AI, o bo h. Finally, pa -
icipan s we e ully deb ie ed abou he s udy's objec i es and
hanked o pa icipa ing.
3.2 | Pa icipan s
Pa icipan s we e ec ui ed ia a uni e si y dis ibu ion lis and
a ious social media pla o ms. A o al o N = 494 pa icipan s
comple ed he s udy. Howe e , N = 200 had o be excluded as
hey ailed he a en ion check a he end o he ques ionnai e
by no co ec ly iden i ying he expe imen al condi ion hey
we e assigned o (see Sec ion3.4 o he a en ion check i em).
The inal sample comp ised N = 294 Ge man- speaking pa ici-
pan s, o whom 60.20% iden i ied hemsel es as emale, 37.76%
as male, and 2.04% as di e se. The a e age age o pa icipan s
was M = 27.14 yea s (SD = 12.08). Mos pa icipan s we e highly
educa ed, holding a uni e si y en ance ce i ica e (63.27%), a
bachelo 's deg ee/p e- diploma (11.56%), a oca ional aining/
app en iceship (9.52%), o a mas e 's deg ee/diploma (7.48%).
Mos o hem indica ed ha hey we e cu en ly en olled in a
s udy p og am (61.22%), ollowed by employees (28.23%) and
eelance s (3.06%).
3.3 | Ma e ials
Fo ou expe imen , we c a ed six di e en email commu-
nica ions. To a y he deg ee o AI- MC, we used an app oach
simila o Liu e al.(2022). Ou manipula ion o se ice c i i-
cali y ollowed a common scena io echnique (see below;
Moza a i e al.2021; Os om and Iacobucci1995; Webs e and
Sunda am 1998). The designed s imulus ma e ial showed an
email communica ion in ol ing a de ec i e p oduc in he con-
ex o a p oduc e u n scena io om an online e aile and was
designed by he esea ch eam based on he s udy design by Liu
e al.(2022). The p oduc e u n scena io was selec ed o i s
ele ance and ealism, mi o ing eal- wo ld cus ome expe i-
ences (see also Huang and Doo son2022). Gi en i s p ominence
in online shopping, we chose o si ua e ou scena io wi hin he
cus ome elec onics sec o (Lohmeie 2024; S a is a Ma ke
Insigh s2024).
The six designed email scena ios illus a ed a se ice employee's
esponse o a cus ome 's e u n eques . Each email acknowl-
edged he cus ome 's eques , exp essed apologies o he in-
con enience due o he aul y i em, and assu ed p omp e und
a angemen by he cus ome se ice. The email concluded wi h
a eques o he cus ome o ollow a p o ided link o comple e
he e u n p ocess and submi pe sonal da a o he e u n p o-
cess (see AppendixA o he s imulus ma e ial). To manipula e
he deg ee o AI- MC, a e he ega ds o he se ice employee,
a sen ence in as e isks was included indica ing he sende o he
email. Fo he condi ion in which he se ice employee w o e
he email (comple e human agency condi ion), we added he
sen ence “This email was pe sonally w i en by he se ice em-
ployee” o he condi ion in which he se ice employee used
AI o w i e he communica ion (sha ed human/AI agency con-
di ion) we added he sen ence “This email was w i en by he
se ice employee using an in elligen au ocomple e sys em (a i-
icial in elligence).”, and o he condi ion in which he AI w o e
he email on behal o he se ice employee (comple e AI agency
condi ion), we included he sen ence “This email was w i en by
an ad anced AI (a i icial in elligence) sys em on behal o he
se ice employee.”
To manipula e se ice c i icali y, pa icipan s we e asked o
ead a sho scena io be o e eading he email communica ion.
Fo he high se ice c i icali y condi ion, pa icipan s we e
asked o imagine ha ing ecen ly pu chased a new and cos ly
lap op om an online e aile . Upon i s i s use, hey disco e ed
ha he compu e was de ec i e in a pa icula ly dis essing
5 o 17
si ua ion o u gen need o he lap op o comple e a c ucial s u-
den assignmen o o p epa e a i al job p esen a ion. In he low
se ice c i icali y condi ion, pa icipan s we e asked o en ision
ha ing ecen ly acqui ed a low- p iced USB cha ging cable on
sale, which hey ound de ec i e upon i s use. Howe e , his
si ua ion was amed as less c i ical, gi en he non- u gen na-
u e o he need o he cable, as hey had o he cha ging cables
a ailable o use.
Excep o he de ec i e p oduc ype and he sen ence indica -
ing he deg ee o AI- MC, he con en and wo ding ac oss he
emails emained iden ical. To imp o e ealism, each email was
p esen ed as a sc eensho om Gmail, a widely used email cli-
en (Pe osyan2024; Rabe2024). As a name o he se ice em-
ployee, Alex Schneide was chosen as bo h he gi en name and
he su name a e among he mos common names in Ge many
(Wikipedia2024; Rüdebusch n.d.). Fo a p e- es o he s imulus
ma e ial, see he Suppo ing In o ma ion.
3.4 | Measu es
T us in he online e aile was measu ed wi h an es ablished
ques ionnai e assessing us in e- comme ce (McKnigh
e al.2002). The ques ionnai e measu es us on he compo-
nen s o bene olence (e.g., “I belie e ha he online e aile
would ac in my bes in e es .”; ɑ = 0.80), in eg i y (e.g., “The
online e aile is u h ul in i s dealings wi h me.”; ɑ = 0.83), and
compe ence (e.g., “The online e aile is compe en and e ec-
i e in p o iding elec onic p oduc s.”; ɑ = 0.81). The in e nal
consis ency o he agg ega e us measu e was ɑ = 0.90. In line
wi h McKnigh e al.(2002), us was addi ionally assessed as
a beha io al in en ion wi h i ems asking pa icipan s o indica e
hei willingness o engage in us - ela ed beha io s, such as
sha ing sensi i e pe sonal in o ma ion (Cu all and Judge1995;
McKnigh e al.2002) wi h he online e aile in he con ex o
his email communica ion. The h ee i ems we used we e as ol-
lows: “I would click on he link in he email.”, “I would p o ide
my con ac de ails (name and add ess) on he linked websi e.”,
and “I would p o ide my bank de ails (IBAN) on he linked
websi e.” As p e ious wo k has p oduced mixed indings e-
ga ding he in luence o gene al us and a i ude owa d AI
when in es iga ing us in AI- MC and in AI (Chua e al.2023;
Liu e al.2022), disposi ion o us was addi ionally measu ed
wi h he i ems o McKnigh e al.(2002; e.g., “In gene al, people
do ca e abou he well- being o o he s.”, “In gene al, mos olks
keep hei p omises.”; ɑ = 0.89.). A i ude owa d AI was assessed
wi h he Ge man e sion o he A i ude Towa ds A i icial
In elligence scale (ATAI scale; Sinde mann e al. 2021). This
sho measu e cap u es a i ude owa d AI wi h i e i ems (e.g.,
“A i icial in elligence will bene i humankind.”, “A i icial in-
elligence will cause many job losses.”, ɑ = 0.81). As a manipu-
la ion and alidi y check, pa icipan s we e asked o indica e
pe cei ed se ice c i icali y (“The as es possible esolu ion o
my se ice eques is c i ical o me.”; Moza a i e al.2021), and
ealism (“This scena io is ealis ic.”) o he p esen ed scena io,
espec i ely. As an a en ion check, pa icipan s we e asked o
indica e how he p esen ed email communica ion was w i en
(“Who w o e he online e aile 's email?”). The a en ion check
u he helped o ensu e ha esponden s unde s ood co ec ly
whe he he email had been w i en by a se ice employee, a
se ice employee wi h he help o an AI o by an AI on behal o
a se ice employee depending on hei espec i e expe imen al
condi ion. Response op ions we e iden ical o he wo ding used
in he expe imen al manipula ion (“This email was pe sonally
w i en by he se ice employee” s. “This email was w i en
by he se ice employee using an in elligen au ocomple e sys-
em (a i icial in elligence)” e sus “This email was sen by an
ad anced AI (a i icial in elligence) sys em on behal o he se -
ice employee.”). All measu es o he s udy we e assessed using
7- poin Like - ype scales (1 = s ongly disag ee o 7 = s ongly
ag ee).
4 | Resul s
4.1 | Da a P epa a ion
The a e age pe cen age o missing da a ac oss all eco ded a i-
ables was 1.80%. These missing da a we e eplaced wi h median
(con inuous a iables) and mode alues (ca ego ical a iables).
4.2 | S a is ical Analyses
ANOVA and ANCOVA analyses we e conduc ed wi h ype III
SS, as expe imen al condi ions yielded unbalanced sample sizes
(Hec o e al.2010; Shaw and Mi chell- Olds1993; see Table1).
To u he examine signi ican main e ec s wi hin ANCOVA,
Tukey pos hoc es s on he es ima ed ma ginal means (EMMs)
o he di e en expe imen al condi ions we e calcula ed.
Acco dingly, o g oup compa isons, EMMs and SE a e epo ed
(H1–H3). E ec sizes a e epo ed in line wi h Cohen(1998).
4.3 | Manipula ion and Validi y Checks
Two- way ANOVAs we e pe o med o e alua e he e ec s o
AI- MC condi ion, se ice c i icali y, and hei in e ac ion on
pe cei ed se ice c i icali y (manipula ion check) and eal-
ism ( alidi y check). Fo pe cep ions o se ice c i icali y, e-
sul s yielded a signi ican main e ec o se ice c i icali y, F(1,
288) = 6.53, p = 0.011, ηp2 = 0.02, while he main e ec o AI-
MC condi ion and he in e ac ion e m we e non- signi ican
(ps ≥ 0.474). We hus u he in es iga ed pe cep ions o se ice
c i icali y among he high and low se ice c i icali y condi ions.
An independen sample Welch's - es e ealed ha in he high
TABLE 1 | Sample sizes ac oss expe imen al condi ions.
Se ice
c i icali y
AI- MC
Se ice
employee
Se ice
employee
assis ed
by AI
AI on
behal o
se ice
employee
High 25 55 74
Low 19 52 69
No e: To al N = 294.
Abb e ia ions: AI- MC, a i icial in elligence- media ed communica ion; AI,
a ici ical in elligence.
6 o 17 In e na ional Jou nal o Consume S udies, 2025
se ice c i icali y condi ion, pa icipan s epo ed highe pe -
cep ions o c i icali y ega ding he esolu ion o hei se ice
eques (M = 6.45, SD = 0.96) han in he low se ice c i ical-
i y condi ion (M = 6.16, SD = 1.03, (284.82) = 2.50, p = 0.014,
d = 0.29 [0.06; 0.52]). Fo pe cei ed ealism, he e we e no sig-
ni ican main e ec s o in e ac ion (ps ≥ 0.439), sugges ing ha
pa icipan s' pe cep ion o ealism did no di e ac oss he six
expe imen al condi ions. A ollow- up one- sample - es agains
he mid- poin o he scale (i.e., 4) indica ed ha pa icipan s pe -
cei ed he p oduc e u n scena ios and email communica ions
as highly ealis ic (M = 5.49, SD = 1.30, (293) = 19.64, p < 0.001,
d = 1.15 [1.00, 1.29]).
4.4 | E ec s o AI- MC on Cus ome T us (H1)
To es H1, we conduc ed a one- way ANCOVA o de e mine
whe he cus ome us a ings signi ican ly di e ed be ween
he h ee AI- MC condi ions. Mean cus ome us a ings se ed
as he dependen a iable, while age, gende , disposi ion o us ,
and he a i ude owa d AI we e en e ed as co a ia es. In line
wi h H1, he main e ec o AI- MC condi ion p o ed signi ican ,
F(2, 286) = 4.65, p = 0.010, ηp2 = 0.03. Rega ding he co a ia es,
disposi ion o us was signi ican ly ela ed o cus ome us
a ings, F(1, 286) = 11.08, p < 0.001, ηp2 = 0.04, while o he co-
a ia es we e non- signi ican (ps ≥ 0.526). Tukey pos hoc es s
e ealed ha he co a ia e adjus ed mean o cus ome us a -
ings was signi ican ly g ea e in he comple e human agency
condi ion (EMM = 5.16, SE = 0.14) han in he sha ed human/
AI agency condi ion (EMM = 4.66, SE = 0.09, (291) = 3.03,
p = 0.007, d = 0.54 [0.185, 0.90]), indica ing a mode a e e ec .
G oup di e ences be ween he comple e human agency con-
di ion and he comple e AI agency condi ion (EMM = 4.84,
SE = 0.08) as well as be ween he sha ed human/AI agency and
he comple e AI agency condi ion ailed o each signi icance
(ps ≥ 0.118). Figu e 1 illus a es he EMMs o cus ome us
a ings ac oss AI- MC condi ions along wi h he SE, 95% CI, and
signi ican g oup di e ences.
As an addi ional es o H1, we conduc ed he same s a is ical
analyses as epo ed abo e wi h he beha io al in en ion us
measu es as dependen a iables ac oss AI- MC condi ions. The
main e ec o AI- MC condi ion was non- signi ican o all h ee
beha io al in en ion us measu es (ps ≥ 0.167). In e ms o he
co a ia es, age and disposi ion o us we e signi ican ly ela ed
o he pa icipan s' willingness o click on he link in he on-
line e aile 's email (age: F(1, 286) = 6.12, p = 0.014, ηp2 = 0.02;
disposi ion o us : F(1, 286) = 6.30, p = 0.013, ηp2 = 0.02), o
sha e hei name and add ess wi h he online e aile (age:
F(1, 286) = 8.46, p = 0.004, ηp2 = 0.02; disposi ion o us : F(1,
286) = 6.26, p = 0.013, ηp2 = 0.02), and o sha e in o ma ion on
hei bank accoun wi h he online e aile (age: F(1, 286) = 4.68,
p = 0.031, ηp2 = 0.02; disposi ion o us : F(1, 286) = 10.07,
p = 0.002, ηp2 = 0.03). O he co a ia es we e no signi ican ly e-
la ed o he beha io al us measu e (ps ≥ 0.053). See Figu e2
o he EMMs and SE wi h he 95% CI o he beha io al in en ion
us measu es ac oss AI- MC condi ions.
Gi en hese indings, H1 p oposing ha cus ome us in an
e- comme ce e aile is highe when he e- comme ce e aile 's
email is labeled as being sen by a se ice employee compa ed
o when he email is labeled as being sen by a se ice employee
assis ed by AI o by AI on behal o he se ice employee could
be pa ly accep ed (see discussion o he in e p e a ion o non-
signi ican inding o he di e ence be ween he comple e
human and he comple e AI agency condi ions and he beha -
io al in en ion us measu es, see Table2 o a summa y o he
s udy esul s).
FIGURE 1 | Cus ome us a ings ac oss AI- MC condi ions (H1). The loa alues abo e each ba ep esen he us sco e (EMM); he e ical
black lines indica e he 95% con idence in e als o he SE. Lines wi h as e isks indica e signi ican di e ences be ween g oups based on adjus ed
p- alues (Tukey pos hoc es s; *p < 0.05; **p < 0.01; ***p < 0.001, wo- ailed).
7 o 17
4.5 | E ec s o AI- MC on he T us Componen s
Bene olence, In eg i y, Compe ence (H2)
To es H2, he same s a is ical analyses as desc ibed o H1
we e conduc ed wi h pe cep ions o he us componen s be-
ne olence, in eg i y, and compe ence as dependen a iables.
Analyses showed a signi ican main e ec o AI- MC condi-
ion on pe cep ion o he online e aile 's bene olence, F(2,
286) = 4.59, p = 0.011, ηp2 = 0.03, in eg i y, F(2, 286) = 4.02,
p = 0.019, ηp2 = 0.03, and compe ence, F(2, 286) = 3.16, p = 0.044,
ηp2 = 0.02. In e ms o he co a ia es, disposi ion o us yielded
signi ican ela ionships wi h all us componen s (bene -
olence: F(1, 286) = 9.53, p = 0.002, ηp2 = 0.03; in eg i y: F(1,
286) = 8.09, p = 0.005, ηp2 = 0.03; compe ence: F(1, 286) = 7.70,
p = 0.006, ηp2 = 0.03). O he co a ia es we e non- signi ican (ps
≥ 0.153).
Tukey pos hoc g oup compa isons on he co a ia e adjus ed
means e ealed ha in he comple e human agency condi ion,
pa icipan s pe cei ed he online e aile as mo e bene olen
and in eg al han in he sha ed human/AI agency condi ion
(bene olence: EMMcomple e human agency = 5.15, SE = 0.18 s.
EMMsha ed hum an/A I agenc y = 4.49, SE = 0.12, (291) = 3.03, p = 0.007,
d = 0.54 [0.19, 0.90]; in eg i y: EMMcomple e human agency = 5.50,
SE = 0.15; EMMsha ed human/AI agency = 5.01, SE = 0.10,
(291) = 2.78, p = 0.015, d = 0.50 [0.14, 0.86]), sugges ing mode -
a e e ec sizes. Fu he , pa icipan s a ed he online e aile
highe on in eg i y in he comple e human agency condi ion as
compa ed o he comple e AI agency condi ion (EMM = 5.09,
SE = 0.08, (291) = 2.42, p = 0.041, d = 0.42 [0.08, 0.76]), in-
dica ing a small e ec size, while di e ences be ween he
comple e human agency and he comple e AI agency condi-
ion we e non- signi ican o bene olence a ings (p = 0.088).
Compa isons o pe cep ions o bene olence and in eg i y
be ween he comple e AI agency and he sha ed human/AI
agency condi ion p o ed o be non- signi ican (ps ≥ 0.337).
Al hough a signi ican main e ec o AI- MC was obse ed
o compe ence pe cep ions, Tukey pos hoc compa isons did
no e eal signi ican di e ences be ween he condi ions (ps
≥ 0.084). Figu e3 shows he EMMs o bene olence, in eg i y,
and compe ence ac oss he AI- MC condi ions, wi h he SE,
95% CI, and signi ican g oup compa isons.
Gi en hese indings, H2 s a ing ha cus ome s' pe cep ion o
bene olence and in eg i y is highe when he e- comme ce e-
aile 's email is labeled as being sen by a se ice employee com-
pa ed o when he email is labeled as being sen by a se ice
employee assis ed by AI o by AI on behal o he se ice em-
ployee could be la gely accep ed (see Table2 o a summa y o
he s udy esul s).
4.6 | Mode a ing E ec o Se ice C i icali y on
he Rela ionship Be ween AI- MC and Cus ome
T us (H3)
To es H3, p oposing ha se ice c i icali y mode a es he e-
la ionship be ween AI- MC and cus ome us such ha insi -
ua ions o high se ice c i icali y, he di e ence in us a ings
is mo e p onounced be ween he AI- MC condi ions compa ed
o si ua ions o low se ice c i icali y, we conduc ed ANCOVAs
wi h he wo be ween- subjec s ac o s o AI- MC condi ion and
se ice c i icali y including hei in e ac ion e m. Mean cus-
ome us a ings, beha io al in en ion us measu es, and
us componen s se ed as dependen a iables; age, gende ,
disposi ion o us , and a i ude owa d AI we e en e ed as co-
a ia es. ANCOVAs e ealed no signi ican in e ac ion be ween
AI- MC condi ion and se ice c i icali y o any o he us
FIGURE 2 | Beha io al in en ion us measu es ac oss AI- MC condi ions (H1). The loa alues abo e each ba ep esen he us sco e (EMM);
he e ical black lines indica e he 95% con idence in e als o he SE.
8 o 17 In e na ional Jou nal o Consume S udies, 2025
measu es (i.e., cus ome us a ings, beha io al in en ion us
measu es, us componen s; ps ≥ 0.559). Simila ly, he main e -
ec s o se ice c i icali y we e non- signi ican o any o he us
measu es (ps ≥ 0.090). The main e ec s o AI- MC and he ela-
ionships be ween he co a ia es and he us measu es we e
consis en wi h p e iously epo ed esul s (see Sec ions4.4 and
4.5). Thus, H3, p oposing ha se ice c i icali y mode a es he
ela ionship be ween AI- MC and us , was ejec ed, sugges ing
ha he pe cei ed c i icali y o se ice si ua ions did no in lu-
ence he e ec s o AI- MC on us as hypo hesized (see Table2
o a summa y o he s udy esul s).
5 | Discussion
5.1 | Theo e ical Con ibu ions
Wi h he widesp ead accessibili y o gene a i e AI, such as
Cha GPT, companies ha e s a ed implemen ing his new ech-
nology in hei wo k lows o ealize cos sa ings and inc ease
e iciency. Speci ically, o ma ke ing and CRM, gene a i e AI
appea s o be p omising (Chui e al. 2023). Applying gene a-
i e AI o acili a e CRM, howe e , aises he ques ion o how
cus ome s eac o i and how AI and cus ome communica-
ion media ed by AI in luence cus ome us – a p e equisi e
o a company's business success (Isae a e al.2020). Wi h he
EU AI Ac , which came in o o ce on 1 Augus 2024, compa-
nies in Eu ope a e obliged o make he use o AI anspa en
o cus ome s (Eu opean Commission2024) so ha unde s and-
ing cus ome s' eac ions owa d he use o AI labeled as his
is essen ial. To add ess his esea ch gap, he p esen esea ch
aimed o un eil how di e en deg ees o AI- MC, co e ing he
spec um o comple e human agency o comple e AI agency, a -
ec cus ome us in an online e aile , he eby ex ending p e-
ious wo k on AI- MC and us in pe sonal ela ionships (e.g.,
Hohens ein and Jung2020; Liu e al.2022, 2023) and online sel -
p esen a ion (Jakesch e al.2019). To gain a deepe unde s and-
ing o he dynamics be ween AI- MC and cus ome us , his
s udy disen angled cus ome us in i s componen s in eg i y,
bene olence, and compe ence and explo ed how pe cep ions o
hese us componen s a e speci ically a ec ed when an online
e aile uses AI- MC in cus ome communica ion. By examining
he us componen s, his esea ch adds o he li e a u e on us
and AI- MC, as mos s udies ha e ocused on he explo a ion o
AI- MC and us on an agg ega e us le el (e.g., Dehghanpou i
e al.2020; Hohens ein and Jung2020; Jakesch e al.2019; Liu
e al.2022, 2023; Longoni and Cian2020). An addi ional goal o
his esea ch was o explo e he ole o se ice c i icali y in he
ela ionship be ween di e en deg ees o AI- MC and cus ome
us , expanding o me esea ch which examined how human-
cha bo in e ac ions (i.e., comple e AI agency) impac cus ome
us in se ice on line se ings (Moza a i e al.2021).
On an agg ega e us le el, he esul s o ou esea ch in-
dica e ha indi iduals a e mo e likely o us an online e-
aile when he CRM communica ion is labeled as w i en by
TABLE 2 | Summa y o he s udy esul s.
Hypo hesis Resul Key inding
H1: Cus ome us in an e- com e aile is highe
when an online e aile 's email is labeled as being
w i en by a se ice employee han when he email
is labeled as being w i en by a se ice employee
assis ed by AI o by AI on behal o he se ice
employee.
Pa ially
suppo ed
• Pa icipan s epo ed highe us in he online e aile
when he email was labeled as w i en by a se ice
employee compa ed o when i was labeled as w i en
by a se ice employee assis ed by AI (mode a e e ec ).
• T us did no signi ican ly di e be ween emails labeled
as w i en by a se ice employee and hose labeled as
w i en by AI on behal o he se ice employee.
H2: Cus ome s' pe cep ions o bene olence and
in eg i y a e highe when an online e aile 's email
is labeled as being w i en by a se ice employee
han when he email is labeled as being w i en by a
se ice employee assis ed by AI o by AI on behal o
he se ice employee.
La gely
suppo ed
• Pa icipan s epo ed highe pe cep ions o he online
e aile 's bene olence and in eg i y when he email was
labeled as w i en by a se ice employee compa ed o
when i was labeled as w i en by a se ice employee
assis ed by AI (mode a e e ec ).
• Pa icipan s epo ed highe pe cep ions o he online
e aile 's in eg i y when he email was labeled as
w i en by a se ice employee a he han by AI on
behal o he se ice employee (small e ec ).
• Bene olence pe cep ions o he online e aile did no
signi ican ly di e be ween emails labeled as w i en by
a se ice employee and hose labeled as w i en by AI
on behal o he se ice employee.
H3: Se ice c i icali y mode a es he ela ionship
be ween AI- MC and cus ome us in an online
e aile , such ha he di e ence in cus ome us
be ween AI- MC condi ions is mo e p onounced
when se ice c i icali y is high compa ed o when
se ice c i icali y is low.
No
suppo ed
• The le el o se ice c i icali y did no signi ican ly
in luence cus ome us le els o di e en AI- MC
deg ees.
15 o 17
Tho mudsson, B. 2023b. Ra e o Gene a i e AI Adop ion in he Wo kplace
in he Uni ed S a es 2023, by Indus y. S a is a. h ps:// www. s a i s a.
com/ s a i s ics/ 1 3 612 51/ gene a i e - a i- adop ion- a e- a - wo k- by- i ndus
y- us/ .
Tu ley, D. 2019. “Human In elligence + A i icial In elligence =
Human Po en ial.” G i i h Jou nal o Law & Human Digni y 2019: 170–
189. h ps:// doi. o g/ 10. 69970/ gjlhd. 6i3. 1043.
Vannes e, B. S., and P. Pu anam. 2021. “A i icial In elligence, T us ,
and Pe cep ions o Agency.” INSEAD Wo king Pape . 2023. 1–46.
h ps:// doi. o g/ 10. 5465/ am . 2022. 0041.
Wal he , J. B., and M. T. Whi y. 2021. “Language, Psychology, and
New Media: The Hype pe sonal Model o Media ed Communica ion a
Twen y- Fi e Yea s.” Jou nal o Language and Social Psychology 40, no.
1: 120–135. h ps:// doi. o g/ 10. 1177/ 02619 27X20 967703.
Webs e , C., and D. S. Sunda am. 1998. “Se ice Consump ion
C i icali y in Failu e Reco e y.” Jou nal o Business Resea ch 41, no. 2:
153–159. h ps:// doi. o g/ 10. 1016/ S0148 - 2963(97) 00004 - 0.
Wikipedia. 2024. “Lis e De häu igs en Familiennamen in Deu schland
[Lis o he Mos Common Su names in Ge many].” h ps:// de. wikip
ed ia . o g / w ik i/ Lis e_ de _ hä u i gs en_ Fam il ienn a men _ in _ D eu s
chland.
Xu, Y., Z. Chen, M. Y.- P. Peng, and M. K. Anse . 2020. “Enhancing
Consume Online Pu chase In en ion Th ough Gami ica ion in China:
Pe spec i e o Cogni i e E alua ion Theo y.” F on ie s in Psychology 11:
581200. h ps://doi.o g/10.3389/ psyg.2020.581200.
Xu, Y., C.- H. Shieh, P. Van Esch, and I.- L. Ling. 2020. “AI Cus ome
Se ice: Task Complexi y, P oblem- Sol ing Abili y, and Usage
In en ion.” Aus alasian Ma ke ing Jou nal 28, no. 4: 189–199. h ps://
doi. o g/ 10. 1016/j. ausmj. 2020. 03. 005.
Yang, Y., Y. Liu, X. L , J. Ai, and Y. Li. 2022. “An h opomo phism and
Cus ome s' Willingness o Use A i icial In elligence Se ice Agen s.”
Jou nal o Hospi ali y Ma ke ing & Managemen 31, no. 1: 1–23. h ps://
doi. o g/ 10. 1080/ 19368 623. 2021. 1926037.
Zanzo o, F. M. 2019. “Viewpoin : Human- In- The- Loop A i icial
In elligence.” Jou nal o A i icial In elligence Resea ch 64: 243–252.
h ps:// doi. o g/ 10. 1613/ jai .1. 11345 .
Zhang, G., L. Chong, K. Ko o sky, and J. Cagan. 2023. “T us in an
AI Ve sus a Human Teamma e: The E ec s o Teamma e Iden i y
and Pe o mance on Human- AI Coope a ion.” Compu e s in Human
Beha io 139: 107536. h ps://doi.o g/10.1016/j.chb.2022.107536.
Suppo ing In o ma ion
Addi ional suppo ing in o ma ion can be ound online in he
Suppo ing In o ma ion sec ion. Da a S1: Suppo ing In o ma ion.
16 o 17 In e na ional Jou nal o Consume S udies, 2025
Appendix A
Scena io Desc ip ion and S imulus Ma e ial
Con ains he scena io desc ip ions and s imulus ma e ials used
in he s udy. The scena ios we e designed o manipula e he de-
g ee o AI agency in CRM communica ion and he le el o se ice
c i icali y in a p oduc e u n con ex . Pa icipan s we e p esen ed
wi h a sc eensho o an email esponse om an online e aile and
asked o imagine hemsel es as he cus ome . The email a ied in
AI agency (comple e human agency, sha ed human/AI agency, o
comple e AI agency) and se ice c i icali y (high s. low). The ollow-
ing sec ion p o ides he exac wo ding o he s imulus ma e ials o
wo selec ed scena ios: (a) a high- c i icali y scena io wi h comple e
AI agency (Figu eA1) and (b) a low- c i icali y scena io wi h com-
ple e human agency (Figu eA2). The emaining scena ios ollow he
same s uc u e and can be in e ed om he me hod sec ion in ou
manusc ip .
High C i icali y Scena io–Comple e AI Agency Condi ion
The ollowing shows you an email communica ion om an online e-
aile o a cus ome . We a e in e es ed in how you pe cei e his email.
Please imagine ha you ha e ecei ed his email you sel om
an online e aile .
Backg ound
You ecen ly pu chased a new and e y expensi e lap op om an on-
line e aile . Upon i s use, you disco e ha he lap op does no wo k.
Howe e , you u gen ly need he de ice o c ea e an impo an p esen a-
ion o you job o you s udies.
Subsequen ly, you con ac ed he online e aile 's cus ome se ice ia
email. This email was w i en by an ad anced AI (A i icial
In elligence) sys em on behal o he se ice employee. You e-
cei e he ollowing eply:
FIGURE A1 | Sc eensho o he high c i icali y scena io–comple e AI agency condi ion.
FIGURE A2 | Sc eensho o he low c i icali y scena io–comple e human agency condi ion.
17 o 17
Low C i icali y Scena io–Comple e Human Agency Condi ion
The ollowing shows you an email communica ion om an online e-
aile o a cus ome . We a e in e es ed in how you pe cei e his email.
Please imagine ha you ha e ecei ed his email you sel om
an online e aile .
Backg ound
You ecen ly pu chased a USB cha ging cable om an online e aile .
The cable was no expensi e, as i was on sale. Upon i s use, you
disco e ha he cable does no wo k. I is no u gen since you ha e
o he cha ging cables a home, bu i is s ill annoying.
Subsequen ly, you con ac ed he online e aile 's cus ome se ice by
email. This email was pe sonally w i en by he se ice em-
ployee. You ecei e he ollowing eply:
