OpenAI's advertising pilot has crossed $100 million in annualised revenue in under two months, the company confirmed this week. That is not a product milestone. It is a structural shift in what ChatGPT is.
The ad business launched in January 2026 with a US pilot. By late March it was working with more than 600 advertisers. According to OpenAI, less than 20 percent of eligible users see ads on any given day, and the company says it has detected no negative effect on "privacy-related trust metrics." What those metrics actually measure, and who is measuring them, OpenAI does not say.
Analysts at Truist estimate OpenAI will bring in under $1 billion in ad revenue in 2026, rising to more than $30 billion by 2030. That figure, if it materialises, would make ChatGPT one of the largest advertising platforms in the world. For context, Twitter's peak annual ad revenue before its acquisition was around $5 billion. OpenAI would more than triple that within four years.
The timing is uncomfortable for a company that has spent years positioning itself as a responsible counterweight to the attention-maximising logic of social media. OpenAI's founding documents describe its mission as ensuring that artificial general intelligence benefits all of humanity. That phrasing does not obviously include selling attention to advertisers. But $30 billion a year has a way of clarifying priorities.
There is a structural argument here that goes beyond the usual complaints about ads being annoying. The advertising model has a specific effect on the systems it funds. Ad-supported products compete for time and engagement. They optimise for click-through, for emotional salience, for the kind of content that keeps users in the funnel. The decisions that maximise ad revenue are often orthogonal to the decisions that produce accurate, helpful, measured responses. Google's search business is the canonical example: eighteen years of ad optimisation produced a results page now widely described as degraded.
OpenAI has said advertising will represent less than half of its long-term revenue. The other pillar is subscriptions: ChatGPT Plus, Pro, and enterprise plans. But the history of hybrid models is not encouraging. When ad revenue accelerates faster than subscription revenue, the incentive structure tilts accordingly. The product team that drives $30 billion in ad revenue gets more engineers than the team optimising for factual accuracy.
None of this is to say the pilot is a catastrophe. The ads so far appear to be conventional display-style placements rather than anything embedded in model responses. OpenAI says it is focused on brand advertising rather than performance advertising, which is a more benign category. And the company's stated red lines, no ads in the system prompt, no altering the model's actual answers to favour advertisers, would be legally enforceable in most jurisdictions if broken. There are guardrails.
But guardrails are not the same as incentives. Guardrails can be relaxed. Incentives compound. The question is not whether OpenAI will honour its stated red lines today; it is what the organisation looks like in 2030 when advertising is a $30 billion line item and the company has shareholders, or public market pressure, or both. That is a different company from the one that published the original safety charter.
The milestone is worth noting precisely because of how fast it arrived. It took Google years to become an advertising company. OpenAI got there in two months. The speed itself tells you something about the gravity of the model: once the infrastructure is in place, ad revenue pulls everything toward it.