Trust and Credability in AI-Generated Product Recommendation
Author
Chahid, Jonas Bruun
Term
4. Semester
Publication year
2026
Abstract
This thesis examines how AI-generated product recommendations influence consumer trust and purchase intention compared with expert-based and peer-based recommendations in digital B2C environments. Drawing on the heuristic decision-making framework, source credibility theory, and research on algorithm aversion, the study employs a quantitative between-subjects experimental design in which 216 respondents are exposed to one of three recommendation sources: AI, expert, or peer. Differences in perceived credibility and trust across sources are analysed using one-way ANOVA, while linear regression is applied to assess the relationships between perceived credibility, trust, and purchase intention. The findings reveal significant differences between all three recommendation sources: expert-based recommendations receive the highest evaluations of credibility and trust, peer-based recommendations occupy an intermediate position, and AI-generated recommendations are evaluated lowest. Peer-based recommendations also generate substantially higher trust than AI recommendations. The regression results show strong positive relationships between perceived credibility and trust and between trust and purchase intention, and all five hypotheses are supported. The study indicates that recommendation source acts as an important evaluative cue for consumers and that credibility and trust are central to their responses to digital recommendations. The comparatively weak evaluations of AI-generated recommendations suggest that firms should not assume AI-based advice will automatically be perceived as credible or trustworthy, but that its effectiveness depends on how the source is presented and reinforced through clear credibility signals, transparency, expert validation, or peer information. The thesis contributes to understanding consumer responses to AI-generated recommendations while acknowledging limitations related to convenience sampling, a single product category, a simulated purchase setting, and the use of purchase intention instead of actual behaviour.
Denne afhandling undersøger, hvordan AI-genererede produktanbefalinger påvirker forbrugernes tillid og købsintention sammenlignet med anbefalinger fra eksperter og andre forbrugere i digitale B2C-miljøer. Med udgangspunkt i teorier om heuristisk beslutningstagning, kildecredibilitet og algoritmeaversion gennemføres et kvantitativt mellemgruppeeksperiment, hvor 216 respondenter udsættes for én ud af tre anbefalingskilder: AI, ekspert eller peer. Ved hjælp af one-way ANOVA analyseres forskelle i oplevet troværdighed og tillid mellem kilderne, og lineær regression anvendes til at undersøge sammenhængen mellem oplevet troværdighed, tillid og købsintention. Resultaterne viser signifikante forskelle mellem alle tre kilder: ekspertbaserede anbefalinger vurderes højest på både troværdighed og tillid, peer-baserede anbefalinger ligger på et mellem niveau, mens AI-genererede anbefalinger vurderes lavest. Peer-baserede anbefalinger skaber desuden markant højere tillid end AI-anbefalinger. Regressionen viser stærke positive sammenhænge mellem oplevet troværdighed og tillid samt mellem tillid og købsintention, og alle fem opstillede hypoteser bekræftes. Undersøgelsen peger på, at anbefalingskilden fungerer som et centralt vurderingssignal for forbrugere, og at troværdighed og tillid er afgørende for deres respons på digitale anbefalinger. De relativt svage vurderinger af AI-genererede anbefalinger indikerer, at virksomheder ikke bør antage, at AI-baserede anbefalinger automatisk opfattes som troværdige eller tillidsvækkende, men at effekten afhænger af præsentationen og støtte gennem tydelige troværdighedssignaler, transparens, ekspertvalidering eller peer-baseret information. Afhandlingen bidrager til forståelsen af forbrugerreaktioner på AI-genererede anbefalinger, men anerkender begrænsninger knyttet til convenience sampling, et enkelt produktkategori-fokus, et simuleret købssetup og brugen af købsintention frem for faktisk adfærd.
[This abstract has been generated with the help of AI directly from the project full text]