The advertising experiment inside ChatGPT has stopped being an American curiosity. On 11 August OpenAI switched on sponsored placements for users in the United Kingdom, Mexico, Brazil, Japan and South Korea, roughly doubling the number of markets where the product carries commercial inventory.
Those five join an initial group of the United States, Canada, Australia and New Zealand. The pace matters more than the list. A pilot that began in the US earlier this year has moved to five additional countries, including two large non-English-speaking markets, within months, which is not the cadence of a company still deciding whether the idea works.
Who sees them
Exposure is confined to logged-in adults on the Free and Go tiers. Anyone paying for Plus, Pro, Business, Enterprise or Education sees nothing, which makes the subscription a de facto ad-removal purchase and sets up a familiar consumer-internet dynamic where the free product subsidises itself.
OpenAI frames the trade differently, presenting advertising as the mechanism that keeps capable models available to people who will never pay for them. Both descriptions are accurate. The tension between them is the thing worth watching.
The trust problem
A chatbot is a harder surface to monetise than a search results page, because the product is a single authoritative-sounding answer rather than a ranked list users are already trained to discount. If a recommendation can be bought, the assistant's core value proposition erodes immediately.
OpenAI's stated answer is separation. Sponsored units are labelled and rendered apart from the generated response, and the company says advertisers cannot influence what the model actually says. It also says conversation content is not exposed to advertisers.
The control surface is unusually detailed for a first-generation ad product. Users can dismiss an individual ad, submit feedback on it, ask why it appeared, and delete their ad data with a single action. There is a personalisation toggle, and an Ads History view listing every ad served with timestamps and brand names, which can be cleared at will.
That last feature is telling. Publishing a complete log of what a user has been shown is not standard practice in digital advertising, and it suggests OpenAI expects far more scrutiny than a typical launch would attract.
Why the geography is interesting
The new markets are not a random sample. Japan and South Korea are among the heaviest ChatGPT-using countries outside North America, and both have advertising industries accustomed to strict disclosure norms. Brazil and Mexico bring scale with much lower willingness to pay for subscriptions, which is precisely the population an ad-funded tier is designed to serve.
The United Kingdom adds a jurisdiction with an active regulator and a strong tradition of complaints-driven advertising enforcement. Testing there early, rather than late, reads as a deliberate choice to surface compliance problems while the format is still malleable.
Outlook
The open questions are commercial rather than technical. Nobody yet knows what an assistant impression is worth relative to a search click, how often users will act on a sponsored suggestion delivered next to a direct answer, or whether advertisers will accept the limited targeting signal that follows from keeping conversations private.
What is already clear is the strategic logic. Inference at consumer scale is expensive, subscriptions convert a small minority of users, and the gap has to be closed somewhere. Advertising is the only mechanism at internet scale that has ever closed it before.
