Study reveals that dissatisfaction with AI might be largely bias-driven

A new survey suggests that consumers may judge an advertisement differently once they learn artificial intelligence made it. The results point to a possible source bias, while also showing that hidden disclosure can create a separate trust problem.

ContentEngine surveyed 446 U.S. adults about an online advertisement for a product they had not planned to buy. Every respondent imagined making the purchase, using the product and feeling satisfied with the result. The survey changed one detail: one group learned before buying that AI had created the advertisement’s words and images, while the other group learned afterward that the ad had used AI without saying so.

The later-discovery group responded less favorably. Their trust in the advertiser fell by 11.7 percentage points, repeat purchase intent fell by 12.6 points and feelings of being misled rose by 13 points. Those differences are important for advertisers, although the survey also leaves open a more complicated possibility: some dissatisfaction may come from the AI label itself, or from assumptions people attach to it.

Among respondents who saw the AI disclosure before purchasing, 76.7 percent said they would trust the advertiser a great deal or a fair amount. The comparable figure for respondents who learned about the AI afterward was 65 percent.

Purchase intent followed the same pattern:

The strongest negative movement concerned feeling misled. Fifty-seven percent of the upfront group said they felt misled a lot or some. That share reached 70 percent among people who learned about the AI after the purchase.

The survey adjusted its statistical tests for eight comparisons using the Holm method. Trust, repeat purchase intent and feeling misled remained statistically significant after that adjustment. The other differences, including regret, recommendation and feeling manipulated, were less conclusive.

That pattern gives advertisers a practical warning. A campaign can produce an immediate sale and a satisfied customer while still damaging future trust when the customer later learns how the ad was made.

A survey of 446 U.S. adults suggests that consumers react more negatively when they learn after a satisfactory purchase that an advertisement was created using AI and that the advertiser did not disclose it. Compared with respondents told upfront, the later-discovery group reported lower trust and repeat purchase intent and felt more misled. The results point to both a disclosure problem and a possible source bias against AI-generated content, but the study cannot separate those effects because its scenarios combined disclosure timing, nondisclosure language and an entirely AI-created advertisement. The practical message for advertisers is to disclose material AI involvement before a decision, identify human accountability and monitor trust beyond immediate campaign performance.

The survey scenario held the product outcome constant. Every respondent was satisfied with the purchase. That detail matters because it removes one common explanation for distrust: disappointment with the product itself.

The remaining difference came from how respondents encountered information about the ad. Some saw the AI disclosure before the decision. Others discovered the information later and learned that the advertiser had left it out. The response gap suggests that the production method became part of the consumer’s judgment, even after the purchase had gone well.

That reaction may reflect source bias. People often evaluate a message partly through their beliefs about its source. A human creative team may signal care, responsibility or accountability. An AI system may signal automation, impersonality or an attempt to influence people at scale. Those impressions can affect trust before anyone evaluates the actual claims in the advertisement.

The survey cannot measure which belief drove the responses. It did not ask respondents why they felt misled or manipulated. It also described an advertisement whose words and images were created entirely using AI. That is a strong scenario. Consumers may react differently when AI helps with image resizing, background removal, translation or draft copy while a human team develops and approves the campaign.

The results therefore support a careful interpretation: a negative reaction may partly reflect a bias against AI-generated content, alongside concerns about disclosure and accountability. The findings do not establish that AI itself caused all of the dissatisfaction.

The two scenarios differed in more than timing. The later-discovery version explicitly told respondents that the advertisement had failed to disclose its AI use. That wording likely made the advertiser appear secretive or careless.

A consumer who hears about AI before buying can include that information in the decision. A consumer who hears about it afterward receives a different message. The issue becomes connected to whether the advertiser withheld information that could have affected the purchase.

This distinction matters because disclosure can carry two meanings:

The survey combines those meanings in the later-discovery scenario. The lower trust scores may therefore reflect a reaction to nondisclosure more than a reaction to AI production. The higher feeling of being misled makes that possibility especially strong.

The upfront group still showed substantial discomfort. Fifty-seven percent felt misled at least somewhat, and 62.3 percent felt manipulated at least somewhat. Those responses suggest that early disclosure can improve trust without removing every negative reaction.

The wording of the hypothetical may explain part of that result. The advertisement persuaded respondents to make an unplanned purchase, and the entire ad had been created using AI. The persuasive act, the surprise purchase and the AI label could each have contributed to the discomfort.

The study is most useful as a warning against treating AI disclosure as a minor compliance detail. Its results do not prove that consumers reject AI, because the later-discovery scenario framed the advertiser as having withheld information and the all-AI setup may have intensified concerns. Still, the consistent demand for disclosure and human review suggests that consumers want both information and accountability. Advertisers should therefore explain when AI materially shaped an ad, say who approved the final work and measure whether the campaign preserves trust after the sale.

A satisfied customer can still question how the customer reached the decision to buy. That is one of the more useful details in the survey.

The scenario gave respondents a positive product experience. It did not ask them to imagine a defective product, an inaccurate claim or a poor customer service interaction. Even with that favorable outcome, many respondents reported feeling manipulated or misled.

This result points toward a distinction between outcome satisfaction and process acceptance. A person may like the product and still care about the methods used to persuade them. The concern can involve the ad’s targeting, its emotional appeals, its claims or the involvement of AI.

The survey only measured the overall reaction. It cannot identify the exact source of discomfort. Future research could separate those factors by testing several versions of the same scenario:

That design would help distinguish source bias from concerns about secrecy, human accountability and persuasive pressure.

Across both groups, 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. A nearly identical 88.3 percent said it was important to know that a person had reviewed and approved an AI-generated ad before publication.

Those findings give advertisers two separate responsibilities. Disclosure tells consumers about the production process. Human review identifies who remains accountable for the final message.

The human-review result was especially consistent. It reached 87.9 percent in the upfront group and 88.8 percent in the later-discovery group. The difference was less than one percentage point.

That consistency suggests that human oversight is not merely a remedy for consumers who feel deceived. It is a broad expectation that applies across different disclosure experiences. Consumers may accept AI as part of the workflow while still expecting a person to check claims, images, targeting and tone.

The European Commission’s explanation of Article 50 transparency obligations provides one regulatory example of how responsibility can be assigned to the organization using an AI system. The IAB AI Transparency and Disclosure Framework takes a risk-based, materiality-driven approach to when and how disclosure should appear in advertising and marketing content.

The survey’s results fit that direction. A label can disclose AI involvement, while an accountable review process gives the label practical meaning.

The survey compares two separate groups of 223 people. Respondents were not randomly assigned to the scenarios, and the samples came from online panel participants. The results were not weighted to represent the U.S. adult population.

Demographic characteristics measured by the survey did not differ significantly between the groups. That helps with comparison, although similar age, gender, income, device and regional profiles cannot rule out differences in attitudes toward technology, advertising or privacy.

The hypothetical setting also limits the findings. Respondents did not view an actual advertisement, spend money or use a real product. Stated trust and future purchase intent can differ from behavior in a real marketplace, where price, brand familiarity and personal need carry more weight.

The all-AI scenario limits the scope in another way. A consumer may view an advertisement differently when AI produces a background image than when AI creates every word and image. The survey does not show how people respond to partial AI assistance.

The wording creates the largest interpretive challenge. The later-discovery group was told that the ad had failed to disclose its AI use. That sentence directly raises the possibility of deception. The 13-point increase in feeling misled is therefore meaningful, though it cannot be treated as a clean estimate of the effect of disclosure timing alone.

These constraints do not make the results useless. They define what the results can support. The findings show an association between disclosure conditions and consumer reactions within these samples. They provide a reason to test the issue with randomized experiments and real campaign data.

Advertisers evaluating AI-generated creative should track more than clicks, conversions and immediate satisfaction. Later trust and repeat purchase behavior may reveal a cost that performance dashboards miss.

A practical review process could include the following steps:

The wording should also be specific. A vague statement that technology helped produce an advertisement may leave consumers unsure whether AI wrote the copy, generated the image, selected the audience or performed a minor editing task. Clearer descriptions allow consumers to judge the relevance of the information.

Advertisers should treat disclosure as part of the consumer experience, not as a footnote added after the campaign is complete. Placement, timing and plain language can affect whether the notice informs a decision or simply explains a decision that has already happened.

The survey also challenges a simple assumption that transparency will settle every objection. Some respondents who knew about the AI upfront still felt misled or manipulated. Brands may need to review the broader persuasive strategy, especially when the ad encourages an unplanned purchase or presents highly personalized content.

The survey’s most useful insight may concern how consumers judge the path to a purchase. Product quality remains important, yet it sits alongside questions about who created the message, whether the advertiser disclosed relevant information and whether a person accepted responsibility for the result.

Some negative reactions may be bias-driven. The AI label can activate doubts about authenticity or intent even when the product works as promised. The later-discovery results also show a separate trust penalty connected to concealment. Both forces can exist at the same time.

For advertisers, that combination calls for more than a disclosure badge. Consumers appear to want meaningful information before they act and evidence that a person remains responsible for the message. A campaign that meets those expectations has a better chance of preserving trust after the sale, including among customers who remain skeptical of AI itself.