ChatGPT Ads: How OpenAI's Ad Test Works and Grows

The most-used consumer AI product in the world is becoming an advertising business. On August 11, 2026, OpenAI expanded its ChatGPT advertising test to five new markets — the United Kingdom, Mexico, Brazil, Japan and South Korea — widening a program that began in the United States earlier this year and had already reached Canada, Australia and New Zealand. Sponsored results now appear inside one of the highest-traffic surfaces in software, and the expansion makes clear that ads are no longer an experiment on the margin of ChatGPT but a core part of its economic model.

What OpenAI is testing

The mechanics are consistent across markets. Ads are shown only to logged-in adult users on the free and ChatGPT Go tiers, OpenAI’s lowest-cost consumer plans. Paying subscribers on Plus, Pro, Business, Enterprise and Education do not see sponsored content — a deliberate line that keeps advertising confined to users who are not otherwise paying for the product.

According to OpenAI’s description of the test, the system decides which ad to show by matching advertiser submissions against the topic of the current conversation, a user’s past chats, and prior interactions with ads. The company’s own example: someone researching recipes might see ads for meal kits or grocery delivery. When multiple advertisers are eligible, OpenAI says it selects the single most relevant one to surface first.

Two design commitments sit at the center of the pitch. Ads are clearly labeled as sponsored and visually separated from the model’s answer, and — critically — OpenAI states that ads do not influence the responses ChatGPT gives. The organic answer, the company says, is optimized only for what is most helpful to the user, with sponsored placements sitting alongside rather than inside the generated text. Whether that separation holds under commercial pressure is the question the entire industry is watching.

The controls users get

The test ships with a set of transparency and privacy features that go beyond what most web advertising offers. Users can dismiss an ad, give feedback on it, and tap to see why a particular ad was shown. They can delete their ad data with a single action and manage ad personalization directly. An Ads History view lists which ads were seen, with timestamps and brand names, and can be cleared at any time.

These controls are not incidental. Regulators in several of the newly added markets — the UK and the EU-adjacent regimes especially — hold strict standards on personalized advertising and data use, and the granular opt-outs read as an attempt to stay ahead of that scrutiny as the test scales internationally.

How the program got here

OpenAI’s move into advertising has been methodical. The company announced in January 2026 that it would begin testing ads for free-tier and ChatGPT Go users in the US, positioning the format as a way to fund free access rather than as a standalone revenue grab. It followed with a self-serve Ads Manager beta on May 5, 2026, opening the platform to US advertisers of all sizes, and added product-feed campaigns for retailers in June. The August expansion to five countries is the next step in a rollout that has moved from a single market to a global footprint in roughly seven months.

The strategic logic is straightforward. ChatGPT reaches more than a billion people, the overwhelming majority of whom pay nothing. That is an enormous audience and, until now, an almost entirely unmonetized one. Reports of leaked internal forecasts have put free-user monetization at roughly $1 billion for 2026 — a figure that would be immaterial next to OpenAI’s costs today but that represents the seed of a business the company clearly intends to grow. Set against OpenAI’s own guidance that it plans to spend on the order of $750 billion on compute through 2030, the appeal of a high-margin advertising line that scales with usage is obvious.

The stakes for the answer engine

Advertising inside a conversational assistant is not the same as advertising inside a search results page, and that difference is why the launch is contentious. Traditional search ads sit above a list of blue links the user still has to evaluate. In ChatGPT, the model is the answer — there is often no list to scroll, no set of competing sources to weigh. An ad placed next to a single authoritative-sounding response carries more implied endorsement than the same ad next to ten ranked links, which raises the stakes on OpenAI’s promise that sponsorship never bleeds into the substance of the reply.

Early signals from the US beta have been read as encouraging by advertisers: reports describe low ad-dismissal rates and no measurable hit to user-trust metrics during the trial. Marketers, unsurprisingly, are split — some see the largest new advertising surface since the rise of mobile, others worry that injecting commercial incentives into a trusted assistant erodes the very trust that makes it useful. That tension is not going away; it is the defining design problem of monetized AI chat.

The expansion also arrives as OpenAI pushes ChatGPT deeper into paid and enterprise territory, from the work-focused “super app” ambitions to an aggressive pricing strategy on its GPT-5.6 model tiers. Ads and cheaper tokens point at the same goal from opposite ends: maximize the number of people using ChatGPT, then monetize them through subscriptions where possible and advertising where not.

What it means

OpenAI has decided that the free tier is an advertising business, and it is building it deliberately. The progression from a US test in January to a five-country expansion in August, complete with a self-serve Ads Manager and retailer product feeds, is not the shape of a tentative experiment. It is the shape of a company standing up a second revenue engine alongside subscriptions — one that scales with the billion-plus users who will never pay a monthly fee. The near-term dollars are small, but the direction is unambiguous.

The separation between ads and answers is the whole ballgame. OpenAI’s entire pitch rests on the claim that sponsorship never touches the model’s response — that the answer is optimized only for the user, with ads sitting cleanly beside it. That promise is easy to keep while the program is small and reputational risk is high. It gets harder as advertiser demand grows and the incentive to blur the line — sponsored “recommendations,” preferred products woven into answers — compounds. How OpenAI polices that boundary will determine whether ChatGPT ads look like a well-labeled utility or like the slow erosion of a trusted tool.

Regulators and rivals are both watching. The choice to launch in the UK, Japan and South Korea — jurisdictions with real teeth on data and advertising — signals confidence, but it also invites scrutiny that a US-only test avoided. And every other assistant maker, from Google’s Gemini to Anthropic’s Claude, now has a live example of what monetizing conversational AI looks like in practice. If OpenAI’s dismissal rates and trust metrics hold up, expect the format to spread fast.

What to watch next: whether ad load stays light and clearly labeled as the program scales, or creeps toward more integrated placements; whether the newly added markets bring regulatory pushback on personalization and data use; and whether OpenAI ever discloses hard revenue figures, which would tell the market how big this business can actually become. The answer engine has found its ad model. The next year decides how much users will tolerate.