Consumer Survey Tests Whether People Would Recommend an AI Advertiser

A survey of 446 U.S. adults found that the timing of an AI disclosure was linked to trust, repeat purchase intent and feelings of being misled. Recommendation intent moved in the same direction, although that result was less conclusive.

ContentEngine tested two versions of a hypothetical online purchase. In one version, the advertisement told the consumer before the purchase that its words and images had been created entirely with artificial intelligence. In the other, the consumer learned about the AI after buying and using the product, even though the purchase had been satisfactory. The difference gave researchers a practical question: would a consumer still want to recommend the advertiser after learning that the ad used undisclosed AI?

The answer leaned toward caution. About 73.1% of respondents who knew about the AI upfront said they would be very or somewhat likely to recommend the advertiser. That figure fell to 67.7% among respondents who learned about the AI afterward. The 5.4 percentage point difference pointed in the same direction as the larger gaps found for trust and repeat purchase intent, although it did not remain statistically significant after the survey’s comparisons were adjusted.

That distinction matters. The survey offers a warning about delayed disclosure, but it does not support a claim that every difference in recommendation behavior was proven. It also shows why brands need to examine more than immediate sales when they evaluate AI-generated creative.

ContentEngine surveyed two separate groups of 223 U.S. adults through SurveyMonkey Audience, for 446 completed responses in total. Each group read one version of a scenario involving an unplanned purchase.

Both versions described an online ad for a product the consumer had not intended to buy. The ad persuaded the person to make the purchase, and the person was satisfied after using the product. Those details held the product experience constant within the scenario.

The timing and wording of the AI information changed:

Respondents then answered eight questions covering trust, future purchase intent, regret, feeling misled, recommendation intent, feeling manipulated, the importance of AI disclosure and the importance of human review.

The setup is useful because it focuses attention on disclosure timing. It also places limits on what the results can show. The later-discovery scenario included an explicit failure to disclose, which may have influenced answers about feeling misled. Since respondents were placed in separate samples rather than randomly assigned within one experiment, differences between the groups could also reflect characteristics the survey did not measure.

A ContentEngine survey of 446 U.S. adults compared reactions to an online advertisement that disclosed entirely AI-generated words and images either before purchase or after a satisfactory purchase. Upfront disclosure was associated with higher trust and repeat purchase intent, while later discovery was associated with more frequent feelings of being misled. Recommendation intent followed the same pattern—73.1% versus 67.7%—but that difference was not statistically conclusive after adjustment. The findings do not forecast every campaign, yet they suggest that advertisers should disclose meaningful AI involvement before consumers act and maintain clear human accountability for the finished work.

Recommendation was one of several measures in which the upfront-disclosure group gave more favorable answers. Among respondents told about the AI before the purchase, 73.1% said they would be very or somewhat likely to recommend the advertiser. Among those who learned about the AI afterward, 67.7% gave that answer.

The difference was 5.4 percentage points. Its 95% confidence interval ran from -13.8 to +3.1 points, and the adjusted p-value was .426. In practical terms, the result showed a directional gap, but the survey could not rule out the possibility that sampling variation or other differences between the groups contributed to it.

Recommendation intent also asks something different from immediate satisfaction. A customer may like the product and still hesitate to attach their own reputation to the company. Recommending a brand carries a small social cost: the person is effectively telling someone else that the advertiser deserves attention or money. A discovery that the ad concealed its production method may make that endorsement feel less comfortable.

The scenario did not ask respondents to explain their answers, so the survey cannot identify the precise reason for the lower recommendation figure. Possible factors include concerns about transparency, doubts about the advertiser’s judgment, discomfort with AI-generated persuasion or the sense that the brand withheld information before the purchase. Those possibilities should remain separate from the measured result.

The recommendation difference was smaller than the gaps on the survey’s two clearest commercial measures.

Among people who received the disclosure upfront, 76.7% said they would trust the advertiser a great deal or a fair amount. In the later-discovery group, 65.0% said the same. That represented an 11.7 percentage point difference.

Repeat purchase intent showed a similar pattern. About 79.8% of respondents told upfront said they would be very or somewhat likely to buy from the advertiser again. The share dropped to 67.3% among those who learned about the AI later, a difference of 12.6 points.

After adjustment for the eight comparisons, both trust and repeat purchase intent remained statistically significant. Their confidence intervals stayed below zero, which means the estimated gap continued to point toward lower trust and loyalty in the later-discovery group under the statistical model used.

Those results give the recommendation finding more context. Consumers may not immediately abandon a brand after learning about AI use. They may still report satisfaction with the product. However, the discovery can weaken the relationship in ways that affect future buying and willingness to speak positively about the advertiser.

That is important for brands that judge creative primarily through short-term conversion rates. The ad in both scenarios worked: it convinced a consumer to make an unplanned purchase. The product also satisfied the consumer. Even with those favorable conditions, later discovery was associated with worse evaluations of the advertiser.

The most useful lesson is not that AI-generated advertising inevitably damages recommendations; this survey cannot establish that. It is that delayed disclosure creates an avoidable trust problem. Consumers can be satisfied with a product and still reassess the advertiser when they learn that AI shaped the persuasive message without telling them beforehand. Brands should therefore treat disclosure timing, human review and relationship metrics as part of campaign design—not as compliance details added after conversion has already occurred.

The strongest measured difference concerned whether respondents felt misled. About 57.0% of the upfront-disclosure group said they felt misled a lot or some. That figure rose to 70.0% when respondents learned about the AI afterward, a 13.0 percentage point difference.

The adjusted p-value was .030, and the 95% confidence interval ranged from +4.1 to +21.9 points. This was the largest gap among the eight measures and one of the differences that remained statistically significant after adjustment.

The wording of the scenario likely matters here. Respondents in the later-discovery group were told that the ad had not disclosed its AI use. That statement directly describes a lack of transparency, so it may have increased the likelihood of a response indicating that the consumer felt misled. The survey cannot separate the effect of learning about AI later from the effect of being told that the advertiser had failed to disclose it.

Still, the practical lesson is clear enough for advertising teams. A disclosure shown after a purchase arrives too late to inform the decision that the consumer has already made. Even if the disclosure explains the process accurately, it cannot give the consumer a chance to consider that information before acting.

Consumers in the upfront group responded more favorably on the main trust and loyalty measures, yet many still reported discomfort. About 57% said they felt misled at least some of the time, and 62.3% said they felt manipulated a lot or some.

That result complicates a simple compliance approach in which a label is treated as a complete solution. The advertisement in the scenario persuaded someone to make an unplanned purchase, and the entire ad was described as AI-generated. Either detail could have affected how respondents viewed the persuasion.

The survey did not ask whether respondents objected to AI itself, the use of AI in advertising, the unplanned purchase or the persuasive techniques in the ad. It also did not compare full AI production with a mixed process involving human writers, designers or editors. A short disclosure may improve transparency while leaving separate questions about claims, targeting, tone and the intensity of the persuasion.

For brands, that means a disclosure should sit inside a broader review process. The useful questions include:

A consumer who sees a clear label can make a more informed choice. That does not automatically make every AI-generated ad appropriate.

Across both survey groups, 90.1% 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: 88.3% said it was very or somewhat important to know that a person had reviewed and approved the ad before publication.

The human-review result was almost identical across the samples. It reached 87.9% in the upfront group and 88.8% in the later-discovery group. That difference was less than one percentage point and was not statistically meaningful.

Disclosure and review answer different consumer concerns. Disclosure tells people how the ad was produced. Review identifies a person or organization that remains accountable for the final claims, images and implications. A label cannot verify that a product claim is accurate. A review process cannot replace the consumer’s right to know that AI played a major role in creating the ad.

The European Commission’s guidance on transparency obligations under Article 50 of the EU AI Act and the IAB’s AI Transparency and Disclosure Framework both reflect the broader move toward clearer responsibility and consumer-facing disclosure. The exact obligations depend on the use case and jurisdiction, but the direction is relevant for any advertiser operating across markets: AI use increasingly needs a documented process rather than an informal internal decision.

The survey does not establish a universal percentage decline in recommendations, and it should not be used as a forecast for every campaign. Its sample was nonprobability-based, the scenarios were hypothetical and the groups were not randomly assigned. The percentages describe the respondents who completed the surveys rather than all U.S. adults.

The findings still support several practical steps:

Brands may also want to record the review behind each AI-assisted ad. That record can identify the tools used, the human approver, the claims checked and the date of approval. Such documentation gives legal, marketing and customer service teams a common account of what happened if a consumer questions the ad later.

The survey’s recommendation result was less precise than its trust, repeat purchase and feeling-misled results. That does not make it irrelevant. Recommendation is a relationship measure, and relationship measures often move more slowly than a single purchase decision.

A consumer may complete a transaction, use the product and feel satisfied while becoming less willing to endorse the company. If that pattern holds in real campaigns, brands could see AI-generated ads produce acceptable short-term results while weakening referrals and future loyalty over time.

The safest reading of the survey is therefore narrow and practical: early disclosure was associated with better consumer reactions than later discovery, while the evidence for recommendation intent was suggestive rather than definitive. Brands that want consumers to trust and recommend them have a strong reason to disclose meaningful AI involvement before the ad influences a purchase and to make clear that a person remains responsible for the finished work.