Research Reveals Purchase Regrets with AI Advertising

Consumers who learn that an advertisement was made with AI after buying a product may carry that discovery into their next brand decision. A new survey links delayed disclosure with lower trust, weaker repeat purchase intent and a stronger feeling of being misled.

A ContentEngine survey of 446 U.S. adults examined what happened when consumers learned about AI involvement at different points in a hypothetical purchase. Every respondent imagined seeing an online ad for an unplanned purchase, buying the product and feeling satisfied with it afterward. One group saw a clear AI disclosure before buying. The other group learned after the purchase that the ad’s words and images had been created entirely with AI and that the ad had not disclosed that information.

The later-discovery group responded less favorably on the measures most closely tied to future customer relationships. About 65 percent said they would trust the advertiser a great deal or a fair amount, compared with 76.7 percent among people told upfront. Repeat purchase intent fell from 79.8 percent to 67.3 percent. Feeling misled rose from 57 percent to 70 percent.

The results support a straightforward business lesson: disclosure timing can affect how consumers judge a brand after the transaction. They also require careful reading. The survey did not prove that delayed AI disclosure alone caused the difference, and it did not show a statistically reliable increase in purchase regret.

The two survey scenarios kept several details constant. In both, the advertised product was something the respondent had not planned to buy. The ad persuaded the person to purchase it, and the person was satisfied after using it.

The timing and wording around AI changed:

That final detail matters. Respondents were not simply asked to compare an ad with a label and an ad without a label. The second scenario explicitly described a failure to disclose. That language could have influenced the large increase in reported feelings of being misled.

Each group included 223 respondents. After reading the scenario, respondents answered questions about trust, future purchases, regret, feeling misled, recommendations, manipulation, AI disclosure and human review.

The survey used top-two response categories. For example, trust combined answers of a great deal and a fair amount. Purchase intent combined very likely and somewhat likely. The analysis adjusted p-values across all eight comparisons using the Holm method.

A survey of 446 U.S. adults suggests that when consumers learn an advertisement was created entirely with AI can influence how they view the advertiser after purchase. Respondents told upfront about the AI use reported higher trust and repeat purchase intent, while those informed afterward were more likely to feel misled. The findings support early, specific disclosure and meaningful human review, but the study’s hypothetical scenarios, separate nonprobability samples and differing wording limit causal conclusions.

Trust showed one of the clearest differences between the groups. Among people told about AI before buying, 76.7 percent said they would trust the advertiser at least a fair amount. Among those who learned afterward, 65 percent gave one of those favorable responses.

That 11.7-point gap had a 95 percent confidence interval from -20.0 to -3.3 points. The adjusted p-value was .040, meaning the difference remained statistically significant after the survey accounted for its eight comparisons.

Repeat purchase intent moved in the same direction. Nearly four in five respondents in the upfront group, or 79.8 percent, said they would be very or somewhat likely to buy from the advertiser again. The share dropped to 67.3 percent among respondents who learned about AI later. The difference was 12.6 points, with an adjusted p-value of .021.

Those results create a problem for advertisers that judge an AI-generated campaign mainly by immediate conversion. The ad in the scenario worked in the short term. It persuaded someone to make an unplanned purchase, and the customer liked the product. The later discovery still weakened the person’s stated willingness to trust the advertiser or buy from it again.

A successful first transaction therefore may conceal a relationship cost. A campaign can generate a sale while leaving a customer less comfortable with the brand’s methods.

The survey deliberately gave every respondent a satisfactory product experience. That design removed one major explanation for dissatisfaction: the product itself failing to meet expectations.

Even with that positive outcome, disclosure timing was associated with different views of the advertiser. The result suggests that product quality and production transparency operate as separate judgments in a consumer’s mind. Satisfaction can describe the product experience while trust reflects how the brand earned the sale.

Purchase regret produced a less certain result. About 53.4 percent of the upfront group said they regretted the purchase a lot or some. The share was 60.5 percent among those who learned about AI afterward, a difference of 7.2 points.

The confidence interval ranged from -2.0 to +16.3 points, and the adjusted p-value was .378. That means the survey did not establish a statistically significant difference in regret after correcting for multiple comparisons. The headline concern around purchase regret is therefore better understood as a warning about possible discomfort and downstream trust, rather than proof that delayed disclosure makes consumers regret every purchase.

The survey also cannot determine whether respondents would have behaved the same way after seeing a real advertisement, spending real money and using a real product. Stated intentions provide useful signals, but actual buying behavior can differ.

Upfront disclosure produced better results on trust, repeat purchase intent and feeling misled, yet it did not make every concern disappear.

Among respondents who knew before buying that the ad was entirely AI-generated:

The scenario itself may explain part of that reaction. Respondents imagined being persuaded to make an unplanned purchase. That kind of persuasive pressure could feel manipulative even when a person knows how the ad was produced. The wording also described an advertisement created entirely with AI, which is a stronger condition than using AI to resize an image, generate alternate headlines or help a human copywriter brainstorm.

The survey did not ask respondents why they felt misled or manipulated. Their answers may have reflected concerns about AI, the persuasive design of advertising, the unplanned nature of the purchase, or several of those factors at once.

This matters for campaign planning. A disclosure can provide information while leaving the underlying creative strategy open to criticism. Brands still need to examine the ad’s claims, emotional triggers, targeting and overall pressure on the audience.

Support for disclosure was high across both groups. Overall, 90.1 percent of respondents said it was very or somewhat important for brands to tell consumers when an advertisement had been created entirely using AI.

Human review received nearly the same level of support. Overall, 88.3 percent said it was very or somewhat important to know that a person had reviewed and approved an AI-generated ad before publication.

The human-review figures were especially consistent. About 87.9 percent of the upfront group and 88.8 percent of the later-discovery group considered human review important. The difference was less than one percentage point, and the adjusted p-value was .768.

Disclosure and review answer different consumer questions. A disclosure explains how the advertisement was made. Human review identifies who remains responsible for its claims, images and implications. A small label cannot replace an accountable approval process.

The IAB AI Transparency and Disclosure Framework V2 takes a risk-based, materiality-driven approach to those decisions. That framework is useful for brands because AI can appear in many parts of production, from copy generation to synthetic images, voice, video and automated personalization. The amount and type of disclosure may need to reflect how central AI was to the consumer-facing content.

European rules also place attention on organizational responsibility. The European Commission’s guidance on Article 50 of the AI Act explains that when an organization uses an AI system under its authority, employees carrying out those instructions are generally not treated as separate deployers. For advertisers, that principle reinforces a practical point: responsibility stays with the business that publishes the work.

A disclosure shown after a purchase can inform a consumer, but it arrives after the decision that transparency was meant to support. Before the purchase, the information can become part of the consumer’s evaluation of the ad and the brand. Afterward, it can feel like a correction to an incomplete interaction.

The difference is especially important for ads that use synthetic people, voices or scenes. Consumers may make judgments about authenticity, endorsement or product experience based on what they see. A later explanation can change their interpretation of the entire ad.

The background research points to a broader visibility problem. A Gallup survey cited in the research found that 79 percent of Americans had seen advertisements in the prior 30 days that appeared to be made using AI. A Canva study reported that 70 percent of consumers believe it will eventually be impossible to tell whether an ad was AI-generated without disclosure.

As synthetic content becomes harder to identify by sight, brands have less reason to treat disclosure as an optional courtesy. Consumers may increasingly rely on the brand’s own statement to understand how a message was produced.

The survey supports several practical steps for brands and agencies:

These steps carry costs. Early disclosure may reduce short-term response among people who prefer human-created advertising. Human review adds time and labor. Clear records can slow fast-moving creative workflows. Those costs are easier to compare with the cost of lost trust when the campaign is evaluated across the full customer relationship.

ContentEngine conducted two online surveys through SurveyMonkey Audience. The surveys were fielded in August 2026 and used separate nonprobability samples of 223 U.S. adults each. Respondents opted into online panels, and the results were not weighted to represent the U.S. adult population.

The groups were not randomly assigned within one experiment. Their measured demographic profiles did not differ significantly on age, gender, household income, device type and major region, but those similarities cannot rule out differences in unmeasured characteristics.

The scenarios also differed in wording beyond timing. The later-discovery group was explicitly told that the ad had failed to disclose its AI use. That wording may have contributed directly to the 13-point increase in feeling misled.

The study examined a fully AI-generated advertisement. Results may differ when AI assists a human creator or handles a limited production task. The survey also did not measure race, ethnicity or education, and it did not ask respondents to explain their answers.

These limitations narrow the claim the data can support. The survey describes reactions among the people who completed it and shows an association between disclosure timing and several consumer attitudes. It does not provide a population-wide estimate or isolate timing as the only causal factor.

The strongest evidence concerns trust, repeat purchase intent and feeling misled. Those measures all moved in the direction brands should take seriously when consumers learned about AI after the purchase. Regret, recommendation and manipulation also moved in unfavorable directions for the later-discovery group, but those differences were less precise and did not remain statistically significant after adjustment.

The broad support for both disclosure and human review gives advertisers a clear signal about consumer expectations. People want information about AI involvement, and they also want a person to remain accountable for the final message.

A campaign that performs well at the point of sale can still create a weaker relationship if the audience later feels that important information was withheld. Early, specific disclosure gives consumers a fair chance to judge the message before acting. Careful human review gives the brand a defensible answer when consumers ask who was responsible for what they saw.