# How do consumers feel about ads inside AI assistants? - URL: https://www.askattest.com/blog/articles/how-do-consumers-feel-about-ads-inside-ai-assistants - Published: 2026-11-09 - Description: Consumers are surprisingly comfortable with ads inside AI assistants – but only when they are obviously an ad. --- Consumers will accept ads inside AI assistants, but only when they are clearly labeled. 41% of AI users say they would be comfortable with an ad marked as one, against 31% uncomfortable. Make the same ad less obvious and the picture inverts: 27% comfortable, 46% uncomfortable. Undisclosed advertising damages trust: many said they would stop using the assistant altogether, and others said they would go back to checking answers elsewhere. The figures here come from Attest’s [2027 Consumer trends report](https://www.askattest.com/our-research/2027-consumer-trends-report), *The mediated consumer*, based on a survey of 4,000 consumers across the US and UK, with 60 AI-moderated interviews run alongside it using [Attest Explore](https://www.askattest.com/attest-explore) . ### 2027 Consumer trends report We analysed 3,252 real purchase journeys to reveal exactly how AI is changing shopper behavior, and what brands need to do in response. [Download report](https://www.askattest.com/our-research/2027-consumer-trends-report) ![](https://www.askattest.com/wp-content/uploads/2026/10/consumer-trends-2027-shape.png) ## **How common are ads inside AI assistants already?** Common enough that this is no longer a question for the future. **41% of AI users say they have already seen an ad inside an assistant.** OpenAI began serving ads inside ChatGPT’s free and Go tiers in the US in February 2026, [rolling out across Europe in August](https://openai.com/index/chatgpt-ads-expands-across-europe/). Google has been placing ads in its AI Overviews, the AI answers above search results, since late 2024. An AI assistant here means a tool you ask questions of in plain language, such as ChatGPT, Gemini, Claude or Perplexity, including the AI answers that appear at the top of a search results page. A marked ad is one identified as paid placement. An unmarked ad is a paid placement presented as part of the assistant’s own answer. ## **What changes when an ad is clearly marked, and when it isn’t?** Comfort roughly inverts. Shown an ad inside an AI answer that is clearly marked as one, **41% of AI users say they would be comfortable and 31% uncomfortable**. Change one thing, an ad that isn’t obviously an ad, and **comfort falls to 27% while discomfort rises to 46%**. The share who are very uncomfortable nearly doubles. What people object to is being advertised to without being told. Comfort with an ad inside an assistant’s answerClearly markedNot obviously an adCompletely fine12%8%Fairly fine29%19%Neither28%27%Uncomfortable20%27%Very uncomfortable11%19%*Base: AI users. Attest 2027 Consumer trends report.* ## **Why does disclosure matter so much with ads in AI assistants?** Because advertisers are buying access to a belief: that the assistant has no stake in the answer. **46% of shoppers who used an assistant during a purchase think its product advice is usually or always impartial.** Only 13% think it rarely or never is. That belief tracks behavior closely: shoppers who think the advice is impartial ruled a product out without checking anywhere else 65% of the time, against 47% among those who think it rarely is. Alongside the survey we ran 60 AI-moderated interviews, 30 in each market. Asked how they would feel about ads appearing in an assistant, people who objected gave almost identical reasons: > *“It would just ruin the whole point of using AI, that it was an unbiased advice.”* > *“I trusted AI a lot because it’s not biased. It’s neutral. So if ads started appearing, I would not trust AI again.”* > *“It would feel like it was no longer unbiased, and it was made on how much a company paid it.”* An advertiser pays to reach people while they hold that belief. The ad is what threatens it. ## **Does seeing ads in AI assistants make people more accepting, or more wary?** More accepting. It’s counterintuitive, but objection is lowest among the people who have actually seen an ad in an assistant. Among those who have noticed one, **17% are uncomfortable with a marked ad and 60% are fine with it**. Among those who have never noticed any, 41% are uncomfortable and 30% fine. The same split holds for unmarked ads: 34% uncomfortable among those who have noticed any ads, against 56% among those who haven’t. That isn’t simply a usage effect. Heavy users do notice more ads, but the gap holds at every level of use. Among people using three or more assistants who have seen an ad, **73% are fine with a marked one**. Familiarity builds tolerance, but only for the format that was disclosed. Nothing here suggests exposure softens people toward ads they weren’t told about. ## **What happens if the disclosure slips?** If shoppers stopped believing the answers were neutral, how would this change their behavior? This is something we dug into during the AI-moderated interviews. Asked how they would react to an assistant being paid to recommend a product without declaring it, the most common answer was leaving: > *“I wouldn’t use Gemini anymore.”* > *“I would never use that AI format again.”* > *“I’d stop using that app, because to me that’s being biased and not giving me the most accurate information.”* > *“I would switch to a different tool, or just not use AI altogether.”* Of the 60 people asked, around 22 said they would stop using the assistant or move to a competitor. The second most common reaction was going back to verifying: > *“I wouldn’t listen to it. I’d maybe check up myself on another platform.”* > *“Instead of treating it like a chatbot, I’d now use it like a search engine.”* > *“I would have to double check everything.”* An assistant is worth advertising against because people act on what it says without looking anywhere else. A shopper who goes back to comparing is a shopper the ad no longer reaches at the moment of decision. These are stated intentions in response to a scenario, not observed behavior, and people routinely overstate how decisively they would act. But the direction is consistent, it came up unprompted in both markets, and it points the opposite way from the survey’s own correlation, where the people who notice the most ads currently check the least. ## **What does “clearly marked” actually mean to a consumer?** The UK’s [CAP Code](https://www.asa.org.uk/type/non_broadcast/code_section/02.html) puts it plainly: marketing communications must be obviously identifiable as such, and must make clear their commercial intent where the context doesn’t. In the US, the [FTC’s guidance on native advertising](https://www.ftc.gov/business-guidance/resources/native-advertising-guide-businesses) treats a promotional message as deceptive when it implies to consumers that it is “independent, impartial, or from a source other than the sponsoring advertiser”. That is a close description of an unmarked ad inside an assistant’s answer. What varies is how prominent it is, and consumers do not agree on what prominent means. Asked what would make an ad acceptable, people were specific and they did not all want the same thing. **A visible label, but no agreement on where.** Some wanted the marker on the recommendation itself. Others wanted the ad separated out of the conversation and shown somewhere else entirely. > *“I think it just needs a box saying sponsored at the top of the advice.”* > *“Maybe a small banner at the bottom of the page.”* **More than one option.** Several people drew the line at an assistant naming a single paid product, and would accept paid placement only inside a comparison. > *“It would have to be multiple ads from multiple different brands, so you would have a choice and not be forced into choosing one brand. Choices is the most important thing.”* **An explanation, not just a label.** A smaller group wanted to know why that product was being shown. > *“I would want full disclosure on why it’s being sponsored and why it’s there. I wouldn’t want it to be a kind of hidden thing.”* **A way to turn it off.** One asked for something no platform currently offers. > *“I would need to be able to ask it to give me results that do not include a paid sponsorship.”* **Relevance to the conversation.** This condition came up repeatedly and isn’t about disclosure at all. An ad connected to what someone is already discussing was widely acceptable. An unrelated one read as an intrusion. > *“If I’m conversing about Pampers with the AI, those paid ads of Pampers can reflect at the bottom of my screen. I don’t want to see football ads.”* > *“If the ad is for something that I’m currently looking for, it’s a good idea. If it’s just something totally off topic, then it’s kind of irritating.”* And for a minority, no format works. Asked what disclosure would make ads acceptable, one answered: *“There isn’t anything that would make them acceptable to me.”* Disclosure is the condition for most people, not for everyone. ### Go further with Attest Explore AI-moderated interviews give you the insights that a survey can’t. Discover where Explore can take you. [Discover Explore](https://www.askattest.com/attest-explore) ![](https://www.askattest.com/wp-content/uploads/2026/11/explore-image-1.png) ## **Which AI assistants show ads, and which don’t?** There is no industry position on ads in AI assistants. The four largest assistants have taken four different ones. - **Perplexity** stopped running ads in February 2026, little more than a year after introducing it. Its ads had been clearly labeled and did not influence what the assistant said. Despite this, the company concluded that being trusted was worth more than the revenue. “A user needs to believe this is the best possible answer,” [one executive told the Financial Times](https://www.ft.com/content/6eec07a5-34a8-4f78-a9ed-93ab4263d43c?syn-25a6b1a6=1). - **Anthropic** [pledged in February 2026](https://www.cnbc.com/2026/02/04/anthropic-no-ads-claude-chatbot-openai-chatgpt.html) that Claude will not carry ads: no sponsored links, no advertiser influence on responses, no third-party product placement. Its stated reason is that advertising would create an incentive to optimize for engagement, when the most useful exchange might be a short one. It has said it would be transparent if it ever revisited that. - **Google** sells ads in AI Overviews and is testing them in AI Mode, but has not put them in the Gemini app. It has also declined to rule them out. DeepMind’s Demis Hassabis said in January 2026 there were [no plans for Gemini ads](https://9to5google.com/2026/01/21/google-reiterates-gemini-wont-have-ads/), but in April the company told investors it was open-minded about them. - **OpenAI** began showing ads in ChatGPT’s free and Go tiers in 2026: launching to US users in February and rolling out across Europe in August. It publishes five advertising principles, one of which is answer independence: ads do not influence what the assistant says, and are always separate and clearly labeled. Perplexity is the case to examine. Its ads were clearly labeled and did not influence the answers, which is the standard consumers in our research say they want, and it withdrew anyway. Compliance was never the problem. The company judged that even disclosed advertising would wear down the reason anyone believes the answer in the first place. OpenAI reads its own data differently. In March 2026 it [reported no negative impact on consumer trust metrics, low ad dismissal rates and improving relevance](https://www.emarketer.com/content/openai-s-chatgpt-ads-hit--100-million-annualized-revenues--doubts-remain), and used those figures to justify expanding the pilot. It also names trust, dismissal and relevance as the three things it is watching as ad load grows. Both sides agree that trust is what makes an assistant worth advertising against. They disagree about how quickly damage would show. ## **What should platforms and advertisers do differently?** **Treat AI advertising as a conversion channel, not an awareness one, and fund it from the budget that matches.** Our data shows AI assistants create a lift for brands the shopper already knew, but do little for ones they didn’t. A brand someone had heard of but never bought wins 24% of first-time purchases without an assistant and 36% with one ([learn how Explore uncovered the mechanism behind this uplift](https://www.askattest.com/blog/research/what-60-ai-moderated-interviews-told-us-that-a-survey-couldnt)). A brand new to the shopper wins about a quarter either way, whether AI is involved or not. So an ad is unlikely to reach anyone who doesn’t already know you, and taking the money from brand building would be paying twice for the same customer. **Take the format where the label is obvious.** When you are offered placement options, the least subtle one tests best with the people it is meant to reach. Media buyers usually prefer formats that blend in. On this evidence, blending in is the problem. **Test the disclosure, not just the creative.** Consumers described at least four different things by “clearly marked”: a label on the answer, a comparison rather than a single product, an explanation of why it appeared, and a way to opt out. Those are design decisions with measurably different consequences, and no platform has settled them. **Decide what happens on the paid tier.** Several people drew a line we didn’t ask about, and it was consistent: ads are the price of a free product, not of a paid one. [And an increasing number of people are paying for AI assistants: around one in five. ](https://www.askattest.com/blog/research/why-gen-ai-is-the-only-subscription-category-still-growing) > *“Seeing as AI has a free option, that would be fine. But for the paid options there should be no ads.”* > *“As long as I’m not paying for the AI, I would have no issue with this.”* **Expect a privacy reaction that search ads don’t get.** People tell an assistant things they would never type into a search box, and assistants remember what they were told. An ad arriving inside that exchange lands differently from one above a results page, and a few people read it as evidence the conversation had been passed on. > *“I would feel my conversations are being leaked, and they are bringing those adverts and sponsors based on my information. I would feel less secure.”* > *“I would be more careful with things I say and do with ChatGPT, because you just never know.”* The second reaction carries a cost for everyone. A consumer who volunteers less gets worse recommendations, and a less useful assistant is worth less to advertise against. One finding runs against all of this. A small group read sponsorship as a quality signal rather than a compromise. > *“For ChatGPT to have such sponsored recommendation, that means it’s from a trusted brand and a genuine brand. It definitely increased my trust.”* Three or four people said versions of that. It is too small a group to build on, but it is a reminder that “consumers hate ads” is not what this data says either. ## **Frequently asked questions** **Are consumers okay with ads in AI assistants like ChatGPT or Perplexity?** Conditionally. 41% of AI users say they would be comfortable with an ad that is clearly labeled, against 31% uncomfortable. Comfort falls to 27% when the ad isn’t obviously an ad. **What happens if an AI ad isn’t clearly disclosed?** Discomfort rises from 31% to 46%, and the share who are very uncomfortable nearly doubles, from 11% to 19%. In follow-up interviews, around 19 of 45 people said they would stop using that assistant or switch to another one. **Does advertising make people trust AI answers less?** The survey can’t settle which way the causation runs. What it shows is that belief in impartiality and willingness to act without checking move together: shoppers who think AI product advice is impartial ruled a product out without checking 65% of the time, against 47% among those who think it rarely is. In interviews, people said undisclosed ads would send them back to verifying elsewhere. **Why did Perplexity stop running ads?** Perplexity dropped advertising in February 2026, little over a year after launch, despite its ads being clearly labeled and not influencing the assistant’s answers. Anthropic has pledged that Claude will stay ad-free, on the grounds that advertising would create an incentive to optimize for engagement. Google has kept ads out of the Gemini app while selling them in AI Overviews, but has not ruled them out for Gemini. **Are frequent AI users more or less tolerant of ads?** More tolerant, but only of disclosed ones. Among people who have noticed an ad in an assistant, 60% are fine with a marked ad against 30% of those who never have. Among people using three or more assistants who have seen one, 73% are fine with a marked ad. **How should platforms disclose ads in AI answers?** Consumers described four different things by “clearly marked”: a label attached to the recommendation, a comparison rather than a single paid product, an explanation of why it appeared, and a way to exclude paid results. Relevance to the current conversation mattered as much as the label itself. ## **Methodology** Figures come from the Attest 2027 Consumer trends report, *The mediated consumer*. We surveyed 2,000 consumers in the US and 2,000 in the UK between 22 and 30 September 2026, with both samples nationally representative on age, gender and region. Questions about advertising inside AI assistants, about impartiality, and about ruling products out were asked only of people who use AI, so those figures are a share of AI users rather than of all consumers. The verbatims come from 60 AI-moderated interviews run through Attest Explore alongside the survey, 30 in the UK and 30 in the US, drawn from the same audience with the same screening applied. Interview responses are stated intentions rather than observed behavior.