Do People Trust ChatGPT Ads? What the Research Shows

In early February 2026, Ipsos found that 63% of Americans think ads would make them trust AI results less. Days later, ChatGPT began showing ads. By the end of August, OpenAI reported more than a billion weekly users and an ad business running at $1 billion a year.

People said one thing and did another. Neither half of that tells an advertiser what it needs to know, which is not whether users will leave ChatGPT over ads, but how they read an ad when it sits underneath an answer they asked for. That is where the research gets useful.

Two campaigns, two trust problems

OpenAI has had a reception problem with advertising in both directions. The first was with its own brand ads. In autumn 2025, research firm System1 tested two of OpenAI’s TV spots for ChatGPT with a panel of US consumers. Both landed in the bottom 20% for long-term brand growth and for short-term sales impact, and most viewers could not identify the brand until the logo appeared in the final seconds, as Futurism reported from Adweek’s coverage. The spots were well made and generic: pleasant stories about AI helping someone, without anything that could only belong to ChatGPT. OpenAI returned to the Super Bowl in February 2026 with a spot for its Codex coding assistant, built around real people rather than AI-generated characters.

The second problem is with the ads it sells. The same Super Bowl carried spots from Anthropic built on the promise that its assistant, Claude, would stay ad-free. BNP Paribas data cited by CNBC showed Claude’s daily active users rose 11% after the game, eMarketer reported. A competitor found it worth paying for Super Bowl airtime to make ads in AI the issue. That alone tells you the trust question is commercially live.

What people say: most expect ads to cost trust

The surveys point the same way on the headline. Beyond Ipsos’s 63%, a survey of US adults by Outward Intelligence found that 41% call AI ads intrusive and nearly half believe ads would undermine a chatbot’s credibility, while 26% find them helpful, its CEO wrote in Fortune. In a May 2026 survey of 1,000 US adults who use AI assistants, Zappi found the biggest concerns were too many ads (42%), answers becoming biased toward advertisers (33%), privacy (32%) and ads being hard to tell apart from real answers (28%).

The same Zappi study also found something the headline numbers hide. When asked to compare, 82% said ads in AI assistants would feel at least as trustworthy as Google Search ads, and a third said more trustworthy. Younger users were far more positive than older ones.

Those findings do not contradict each other. Asked about ads in AI in the abstract, people imagine the worst case: an assistant that quietly recommends whoever paid. Asked to compare a labeled AI ad with a search ad they already accept, the fear shrinks. The concern is specific. It is about ads leaking into answers, not about ads existing.

What people do: almost nobody pays to leave

Behavior tells a calmer story. In Outward Intelligence’s survey, only 8% said they would pay monthly for ad-free AI, and 40% chose a free version with ads. Zappi found that 60% of free users would keep using an ad-supported version, while 17% would pay to remove ads and 17% would switch to an ad-free assistant. Usage has not collapsed either: by August, OpenAI reported more than a billion weekly users.

So the risk is not an exodus. It is quieter. People who do not leave can still discount what they read. Perplexity, which is ending its own ad program, made this argument: one executive warned that ads risk making users “suspicious of everything,” Campaign reported. And Zappi found another defensive reaction: 52% said they would change how they use memory and personalization if ads were introduced, with 27% saying they would become less willing to use personalization at all. Users protect themselves by telling the assistant less, which makes both the answers and the ads less relevant.

The incentive behind that worry is real, and it is the subject of why a subscription-funded assistant ended up selling ads. OpenAI’s answer is a set of rules: ads labeled, kept separate from answers, excluded near health, mental health and politics, and not shown to accounts identified as belonging to users under 18. Whether users believe those rules is what the surveys are measuring.

What the trust gap means if you advertise there

For an advertiser, the doubt users carry lands on your ad. Someone who suspects answers might be bought reads the sponsored card underneath with that suspicion switched on. Three things follow.

First, be relevant and obviously an ad. Copy that picks up the topic of the conversation and says plainly what you offer works with the format. Copy that imitates the assistant’s voice or poses as advice confirms exactly what users fear. The 28% who worry ads will be hard to tell apart from answers are the people you lose by blurring the line.

Second, your brand does more work in a small space. A ChatGPT ad gives you a name, a favicon, a title, a line of copy and an image. If none of those is recognizably yours, you are a stranger asking for trust in a setting where trust is low. OpenAI’s own TV spots showed what happens when brand cues arrive too late. Making sure your name, colors and visual style are recognizable at a glance is the work of brand identity strategy built on assets people recognize.

Third, measure the trust cost rather than assuming it. Click-through rate will not show it. Compare conversion rates and order values from ChatGPT traffic with your search traffic, watch whether branded search rises or falls, and use holdout tests where you can, the approach behind performance marketing that measures what ads actually caused. What is available to buy, and whether a test is worth it at all, is covered in the buyer’s guide to ChatGPT ads.

People have not rejected ads in AI. They have put them on probation. The advertisers who do well there will be the ones whose ads look like ads, come from brands people recognize, and do nothing to confirm the suspicion users arrive with.

Frequently Asked Questions

Do people trust ads in ChatGPT?

Most say ads in AI tools would lower their trust. Ipsos found in February 2026 that 63% of Americans think ads would make them trust AI results less. But a May 2026 Zappi survey of AI assistant users found 82% said AI ads would feel at least as trustworthy as Google Search ads. The concern is specific: people worry that advertising will influence answers or be hard to tell apart from them, rather than objecting to labeled ads as such.

Will people pay to avoid ads in ChatGPT?

Few say they will. An Outward Intelligence survey found only 8% of US adults would pay monthly for ad-free AI, and Zappi found 17% of free users would pay to remove ads, while 60% would keep using an ad-supported version. ChatGPT itself lets Free users choose an ad-free mode with fewer daily messages, and paid Plus, Pro and Business plans carry no ads.

How did OpenAI’s own ChatGPT ad campaign perform?

When research firm System1 tested two of OpenAI’s autumn 2025 TV spots for ChatGPT with US consumers, both ranked in the bottom 20% for long-term brand growth and short-term sales impact. Most viewers could not tell the ads were for ChatGPT until the logo appeared at the end. The ads were well produced but generic, without distinctive brand cues early in the spot.

What are the main challenges of advertising on ChatGPT?

Beyond practical issues such as eligibility, targeting and measurement, the main challenge is trust. Users are wary that ads could bias answers, and many say they would share less with the assistant if ads used their data. That makes clear labeling, recognizable branding and honest copy more important than in search. Advertisers should also measure downstream conversions rather than clicks, since doubt shows up after the click, not before it.