AI Ad Research Raises a Customer Comfort Question as 14% Would Feel No Manipulation at All

A successful purchase can still leave a customer uneasy about the advertiser. New survey findings suggest that when people discover AI involvement after the fact, trust and repeat purchase intent suffer even when the product performs well.

A ContentEngine survey of 446 US adults examined how consumers reacted to an online advertisement created entirely with artificial intelligence. Every respondent imagined buying a product they had not planned to purchase and feeling satisfied with it afterward. The main difference was timing: one group saw an upfront AI disclosure, while the other learned about the AI later and was told the advertisement had not disclosed it.

The later-discovery group gave the advertiser weaker scores across the measures that matter most after a sale. Trust fell by almost 12 percentage points, repeat purchase intent fell by nearly 13 points, and the share who felt misled rose by 13 points. The results point to a practical concern for brands: disclosure affects the relationship around the purchase, even when the purchase itself turns out well.

The headline’s 14% figure raises a useful question about how comfortable customers actually feel with AI-generated persuasion. The survey materials supplied for this analysis report the manipulation result by combining the two strongest negative responses: feeling manipulated a lot or some. They do not provide the separate percentage for respondents who selected not at all.

That distinction matters. In the group told about AI upfront, 62.3% said they felt manipulated either a lot or some. In the later-discovery group, the figure was 70.4%. The remaining respondents selected either not too much or not at all, but the published table does not divide those two categories.

The exact share who felt no manipulation at all therefore cannot be confirmed from the reported top-two figures alone. The headline still captures the consumer comfort problem: a relatively small portion of the audience may feel completely at ease with an AI-generated advertisement, while many others retain at least some concern. Brands evaluating these campaigns should pay attention to the full response distribution rather than treating the least negative category as the only sign of acceptance.

A survey of 446 U.S. adults suggests that AI advertising can affect the customer relationship even when the advertised product performs well. Respondents who learned about an ad’s AI involvement after making a satisfactory purchase reported lower trust, weaker repeat purchase intent, and a stronger feeling of being misled than those told upfront. The findings do not prove that disclosure timing alone caused the differences, but they support a practical approach: disclose material AI involvement before consumers act, explain it plainly, and keep a person accountable for the final claims and creative.

Among people who saw the disclosure before buying, 76.7% said they would trust the advertiser a great deal or a fair amount. That share dropped to 65% among people who learned about the AI afterward.

The difference was 11.7 percentage points. After adjustment for the survey’s eight comparisons, the result remained statistically significant. Its 95% confidence interval ranged from a 3.3-point decline to a 20-point decline.

The finding carries a straightforward commercial implication. The advertisement could still persuade a customer, and the product could still meet expectations, while the brand lost some credibility once the production method became known. A customer may separate satisfaction with the item from confidence in the company that promoted it.

Trust also has a longer shelf life than a single conversion. It can affect how people interpret later claims, how much scrutiny they apply to future campaigns, and whether they give the brand another opportunity. A disclosure placed after the transaction reaches the customer after the trust judgment has already been made.

The survey found a similar gap in future buying behavior. In the upfront-disclosure group, 79.8% said they would be very or somewhat likely to buy from the advertiser again. The figure was 67.3% among respondents who learned about the AI later.

That 12.6-point difference was statistically significant after the multiple-comparison adjustment. It suggests that disclosure timing may matter beyond a consumer’s immediate reaction to creative work. The later discovery created a weaker basis for the next purchase, even though the first product had delivered a satisfactory outcome.

A repeat purchase decision involves more than product quality. It also reflects the customer’s view of the seller’s judgment and candor. When AI use is hidden, the customer may begin to wonder what else the brand has chosen not to explain. The survey did not measure that specific thought, but the lower trust and purchase intent are consistent with that kind of concern.

For advertisers, the result creates a measurement warning. A campaign dashboard may show a strong click-through rate, conversion rate, or initial sales result. Those figures cannot reveal whether the audience would feel comfortable buying from the same brand again after learning how the creative was made.

The central lesson is not that consumers reject every AI-assisted advertisement. It is that brands should not treat a successful conversion as proof that the audience is comfortable with how the ad was made. Early disclosure, meaningful human review, and testing of consumer reactions give advertisers a better chance to protect trust beyond the first sale. The survey has important limits, including separate nonprobability samples and a hypothetical scenario, but those limits make the findings a reason to test transparency—not a reason to ignore it.

The strongest measured gap involved feeling misled. Seventy percent of respondents who learned about AI after the purchase said they felt misled a lot or some. Among those who received the disclosure before buying, 57% gave one of those responses.

The 13-point difference remained statistically significant after adjustment. It was the largest gap in the survey’s table.

The product outcome was held constant in the scenario. Every respondent was told that the purchase was satisfactory. That design helps show how information about the advertisement can shape the relationship even when the customer has no stated complaint about the product itself.

The wording also needs careful interpretation. The later-discovery version explicitly said that the advertisement had failed to disclose its use of AI. That phrase may have prompted respondents to focus on nondisclosure and may have increased the feeling of being misled. The study therefore shows an association between the scenario and the response. It does not isolate disclosure timing as the sole cause.

The upfront group responded more favorably on trust, repeat purchase intent, and feeling misled. The disclosure still left substantial discomfort in the scenario.

Among those informed before buying:

Several features of the scenario may explain the unease. Respondents imagined an advertisement that persuaded them to make an unplanned purchase. That persuasive setup could itself feel manipulative, regardless of the production method. The scenario also described an advertisement whose words and images were created entirely with AI, which is a stronger use case than AI-assisted editing, resizing, or copy suggestions.

The survey did not ask respondents to explain their answers. That leaves several possibilities open. Some people may object to automated creative work. Others may dislike the persuasive pressure of an unplanned purchase. Some may want a human to remain accountable for the claims and images. The data support concern about consumer comfort, while the reasons behind that concern require more research.

The survey found broad support for transparency across both scenarios. About 90.1% of all respondents said it was very or somewhat important for brands to tell consumers when an advertisement had been created entirely using AI.

Human oversight received nearly the same support. About 88.3% said it was very or somewhat important to know that a person had reviewed and approved an AI-generated advertisement before publication.

The consistency of the human-review response stands out. It was 87.9% in the upfront-disclosure group and 88.8% in the later-discovery group, a difference of less than one percentage point. The timing of the scenario barely changed that preference.

Disclosure and review answer different consumer questions:

A label alone may tell consumers that AI was involved. It does not tell them whether anyone checked for unsupported claims, strange visual details, discriminatory targeting, or an implication that a human creator would have caught.

Early disclosure gives consumers information before the decision. That timing allows the audience to judge the advertisement with the production method in view. It also reduces the chance that customers later experience the label as a correction or a reveal.

A workable process for advertisers could include:

The right level of detail will vary by format and use. A short notice may be appropriate for a display ad, while a campaign landing page may need a fuller explanation. A vague phrase such as AI enhanced could leave customers unsure whether AI generated a minor variation or the entire message.

These mistakes can make an AI advertising campaign look successful while weakening the relationship around the purchase.

The survey’s results align with a wider movement toward more specific AI transparency practices. The Interactive Advertising Bureau’s AI Transparency and Disclosure Framework V2 uses a risk-based, materiality-driven approach for deciding when and how AI involvement should be disclosed in consumer-facing advertising and marketing content.

That approach gives brands room to distinguish between low-impact production assistance and content that substantially shapes what the audience sees or believes. It also places attention on the consumer-facing effect, rather than treating every software feature as equally important.

European Commission guidance on Article 50 of the EU AI Act addresses transparency obligations and identifies the legal person operating under whose authority an AI system is used. In an advertising setting, that can include the company directing the work. Individual employees working under that company’s instructions are not treated as separate deployers under the cited guidance.

The legal requirements will depend on the jurisdiction, the type of system, and the role AI plays in the content. Brands should obtain qualified legal advice for specific campaigns. The broader direction is clear enough for a practical policy: companies remain responsible for what reaches the audience, even when software performs much of the production work.

The findings are informative, but they should not be treated as a population estimate for all US adults. Each tranche included 223 respondents from a nonprobability online panel. The two groups were separate samples rather than people randomly assigned to one of two versions of the same experiment.

Measured demographic characteristics did not differ significantly between the samples, but that does not rule out differences in unmeasured traits. The later-discovery scenario also included an explicit statement that the advertisement had failed to disclose AI use. That wording may have affected the results directly.

The scenario was hypothetical. Respondents did not see a real advertisement, spend money, or use a real product. Stated intent can differ from behavior. The survey also examined advertisements created entirely with AI, so the results may not apply to a campaign where AI only helped a human writer produce variations or assisted with image resizing.

These limits reduce the strength of broad claims about every AI advertisement. They do not remove the practical lesson from the comparison. Timing, wording, and accountability deserve testing before a campaign reaches a large audience.

Campaign teams often focus on performance metrics that arrive quickly. AI-generated creative makes several slower measures worth tracking as well.

Useful questions include:

The most useful testing may compare several disclosure placements and wordings before launch. It can also separate different types of AI use. A consumer response to fully generated copy and images may differ from a response to automated background removal or spelling assistance.

That level of testing helps a brand avoid a narrow definition of success. A campaign that produces an immediate sale while weakening future trust may look efficient in a short report and expensive over time.

The survey presents disclosure as part of the customer experience rather than a technical footnote. Consumers broadly support knowing when AI created an advertisement, and they place nearly equal value on human review. Later discovery was associated with lower trust, lower repeat purchase intent, and a stronger sense of being misled.

The headline’s focus on the small group that feels no manipulation at all points to a deeper issue: customer comfort is uneven, and a disclosure cannot guarantee acceptance. Brands still need to judge the claims, persuasion tactics, targeting, and quality of the creative.

For companies using AI in advertising, the safest operating assumption is simple. Tell consumers before they act, explain the role AI played in plain language, and keep a person accountable for the final message. That approach gives the customer a fair chance to decide what the advertisement means to them and gives the brand a stronger basis for trust after the purchase.