AI chatbots may build more consumer trust than humans in luxury sustainability messaging

When a luxury fashion brand tells shoppers about its sustainability efforts, who delivers the message may matter as much as what the message says. A new scenario-based experiment, published in Fashion and Textiles, finds that consumers judged identical sustainability information as more trust-building when it was attributed to an artificial intelligence chatbot rather than a human agent, but only for one specific quality of the information: how objective it seemed. The study, led by Yujeong Won and colleagues with senior author Yuri Lee, analyzed responses from 346 U.S. consumers and suggests that the perceived source of a message can quietly reshape how shoppers evaluate a brand’s environmental claims.

The research team recruited 360 participants aged between their twenties and forties through Amazon Mechanical Turk and, after screening out invalid or inattentive responses, retained 346 in the final sample. Each participant was randomly assigned to one of two conditions. In both, they viewed a simulated online chat with a luxury brand service provider named Alex, who presented a sustainable handbag and then supplied information about the brand’s environmental initiatives. The message content, tone, and structure were identical across conditions; the only difference was whether Alex was presented as a human service representative or as a robot character. The scenario used Gucci as the focal brand, with claims adapted from sustainability commitments publicly reported by Gucci and its parent company Kering in 2020 and 2021, including a target to reduce environmental impact by 50 percent by 2025 and a commitment to regenerate one million hectares of land while protecting critical habitats.

Rather than manipulating the message itself, the researchers measured two post-exposure perceptions of information quality. Perceived objectivity captured the extent to which participants saw the sustainability information as factual, unbiased, and credible, while perceived persuasiveness captured how clear, compelling, and convincing it appeared. Participants then rated their brand trust, their intention to accept the information, and their intention to purchase the sustainable product. The data were analyzed using partial least squares structural equation modeling in SmartPLS 4.1.0, with 5,000 bootstrap samples for significance testing and bootstrap-based multi-group analysis to compare the two provider conditions.

The core results confirmed both hypothesized pathways. Perceived objectivity had a significant positive effect on brand trust (β = .291, p = .008), and perceived persuasiveness had an even stronger one (β = .392, p < .001). Brand trust, in turn, was strongly associated with both information acceptance intention (β = .552, p < .001) and purchase intention (β = .570, p < .001). Bootstrapping of indirect effects showed that both information qualities influenced the two behavioral intentions through brand trust, positioning trust as the psychological bridge between how consumers judge a message and what they intend to do about it.

The most striking finding emerged when the researchers compared the two provider conditions. The indirect effects of perceived objectivity on information acceptance and purchase intention through brand trust differed significantly by provider type, and they were significant only when the message was attributed to the AI chatbot. For the human-agent condition, these indirect paths were not statistically significant. By contrast, the indirect effects of perceived persuasiveness through brand trust were significant and statistically indistinguishable across both provider types. Because the message content was held constant, the authors interpret this as a source-cue effect: consumers did not receive objectively different information from the chatbot, but they perceived the same information as more objective when an algorithmic source delivered it, consistent with the machinery heuristic from Sundar’s MAIN model, under which people treat machine-generated content as more neutral and less emotionally biased than human communication.

The study also examined perceived brand luxury, a multidimensional construct covering quality, extended self, tradition, and hedonic value, as a possible boundary condition. Splitting the sample at the mean score of 5.671 produced a low-PBL group of 165 participants and a high-PBL group of 181. In the low-PBL subgroup, the strength of the link between brand trust and both behavioral intentions differed significantly by provider type, with trust translating more strongly into intentions when the information came from the AI chatbot. In the high-PBL subgroup, no significant provider-based differences appeared. The authors are careful to frame these results as subgroup-based evidence rather than a formally tested interaction, because the procedure compared provider conditions within each subgroup without directly testing whether those differences varied between subgroups.

Several technical caveats accompany these findings. Discriminant validity was problematic for some construct pairs: the heterotrait–monotrait ratio exceeded the conservative threshold for four pairs, with point estimates above 1.0 for information acceptance versus purchase intention and for perceived persuasiveness versus purchase intention, meaning these closely related constructs could not be fully separated empirically. Cronbach’s alpha fell below conventional thresholds for perceived objectivity (.519) and purchase intention (.559), although composite reliability values of .757 and .771 exceeded the recommended level. The model also showed only low out-of-sample predictive power relative to a linear regression benchmark, with lower root mean squared error than the benchmark for just two of six indicators of the key target constructs. The authors note that the model should be read as primarily explanatory rather than predictive.

The practical implications are nuanced. The findings suggest that AI chatbots are not universally superior communicators for luxury brands. Their apparent advantage lies specifically in strengthening perceptions of objectivity, which matters most when consumers are skeptical of greenwashing and demand verifiable claims. Perceived persuasiveness, by contrast, worked equally well regardless of who or what delivered the message. The authors suggest that brands whose luxury image is not yet firmly established in consumers’ minds may benefit most from AI-based sustainability communication, while brands with strong established luxury perceptions may see little difference between providers. They advise managers to pilot-test AI-based sustainability messaging with their own target consumers before broad deployment, given the subgroup-based nature of the evidence.

The limitations are substantial and acknowledged in detail. The experiment used a single brand, Gucci, so results may not generalize to other luxury houses with different brand equity or sustainability visibility. The scenario-based design cannot capture the nuances of real-world consumer–chatbot interactions, and the two conditions differed in visual presentation as well as provider type, meaning the observed effects may reflect anthropomorphism or social presence cues rather than provider type alone. Neither message objectivity nor persuasiveness was directly manipulated at the stimulus level, so their causal roles remain to be established. The sample was limited to U.S. consumers and was not restricted to experienced luxury purchasers, and dichotomizing perceived brand luxury at the mean reduced statistical power, with cell sizes ranging from 77 to 94 participants. Cross-cultural replication and formal interaction testing with continuous moderators are recommended next steps.

For a sector under intense scrutiny over environmental claims, the study offers a measured takeaway: the messenger is not a neutral pipe for information but an active cue that shapes how consumers judge its credibility. When luxury brands communicate sustainability through an AI chatbot, shoppers may extend the message a degree of algorithmic neutrality that a human agent does not receive, and that perceived neutrality appears to matter most for translating objective-seeming information into trust, and trust into intentions to accept the message and buy the product. Whether that pattern holds in real conversations, across brands and cultures, and for consumers with very different relationships to luxury remains an open question that the authors’ own subgroup evidence only begins to answer.

Subject of Research: How AI chatbot versus human provider cues shape consumer trust and behavioral intentions in luxury fashion brands' sustainability communication

Article Title: Luxury fashion brands’ AI-powered sustainability chatbot communication: how information qualities and brand luxury perception shape consumer trust and behavior

Article References: Won, Y., Jang, S., Sim, Y., Lee, Y., & Lee, Y. (2026). Luxury fashion brands’ AI-powered sustainability chatbot communication: how information qualities and brand luxury perception shape consumer trust and behavior. Fashion and Textiles, 13(1), Article 43. https://doi.org/10.1186/s40691-026-00493-8

Image Credits: AI Generated

DOI: 10.1186/s40691-026-00493-8

Keywords: artificial intelligence, chatbots, luxury fashion, sustainability communication, brand trust, consumer behavior, perceived objectivity, machinery heuristic, PLS-SEM, purchase intention, greenwashing, perceived brand luxury