Two hundred days. That is the distance between OpenAI’s first sponsored placement inside ChatGPT and a billion dollar annualized run rate.
In April, six weeks into the pilot, the figure was $100 million annualized. Roughly ten times in four months.
OpenAI is targeting $2.5 billion in ad revenue for 2026, and Forbes reported the milestone this week alongside the market expansion.
I covered this back in February, when ads first appeared in ChatGPT and the entire industry argued about whether the answers would stay independent.
That was the wrong argument to have. The answers were never the exposed part. The media plan was.
The ramp is faster than the ad market is used to
Distribution explains most of it. Ads run on the Free and ChatGPT Go tiers, which together account for roughly a billion weekly active users, while Plus, Pro, Business, Enterprise and Education stay ad free.
So the inventory is the largest free consumer audience assembled since mobile social, and it arrived with intent already attached.
Nobody opens ChatGPT to browse.
That is the part media buyers keep underrating. Search intent has to be inferred from three words in a box.
In ChatGPT the user has usually described their budget, their constraints and their last failed attempt before any commercial moment appears in the conversation. The targeting signal is not a keyword. It is a briefing.
Ads are now live in more than 40 countries. The rollout ran the United States in February, the Ads Manager in May, then the United Kingdom, Mexico, Brazil, Japan and South Korea in August, then 31 European markets from 24 August.
Six months, four continents. That pace is not a product decision, it is a financing decision, and it tells you how central this line is expected to become.
What actually changed this week is the price of entry
Self-serve buying opened across India, Europe, the Middle East and North Africa. India’s self-serve launch is set for 4 September, with a daily minimum around ₹725. About $7.60.
Seven dollars and sixty cents. That is the floor.
OpenAI says small and medium businesses already make up a material share of its advertisers, which at that entry point is not surprising. The auction does not fill with brands. It fills with everyone.
And that sets a clock.
Every ad platform has a window where inventory is underpriced because demand has not caught up with supply. Facebook’s window lasted years. TikTok’s lasted months.
This one is being priced by buyers who watched both happen and are not going to wait politely for a case study.
If you are planning to test this in Q1, be clear about what you are actually deciding. You are choosing to enter after the people who tested in Q3 have trained the auction, built creative libraries, and learned which query shapes convert.
That gap is not a small disadvantage. On every previous platform it was the difference between acquiring customers below market and acquiring them at market for the following three years.
The channel that took your clicks is selling them back
Here is the part that should sit uncomfortably for anyone who owns a content budget.
For two years the conversation in marketing has been about AI answers absorbing traffic that used to reach your site. Synthesis without citation, citation without clicks, and a set of dashboards that stopped meaning what they used to mean.
The same interface now offers to sell you placement inside those answers.
This is not a scandal. It is the exact arc Google ran on organic search across fifteen years, compressed into roughly eighteen months.
The compression is the story, not the mechanic. You do not get a decade to adjust your mix. You get the next two or three planning cycles.
It is also the second half of the shortlist economy problem. If the assistant builds the shortlist, there are exactly two ways onto it, and now both are purchasable in different currencies.
So treat it as two separate budget lines, because they solve different problems.
The first line buys presence in the answer. Sponsored placements are labelled and visually separated from the organic response, which is good for user trust and useless for the thing most brands actually lost. A paid slot does not make the model cite you when nobody is paying.
The second line earns presence in the answer. That is visibility work, it runs on a timeline measured in months rather than in bid adjustments, and it compounds. If you want to know where your brand currently stands inside AI-generated answers before you spend anything, GEOflux.ai (geoflux.ai) was built for that measurement, mapping not just whether you are mentioned but why.
Confusing the two lines is the expensive mistake available this quarter. Paid placement is a lever you rent by the day. Earned presence is an asset, and it does not appear in the Ads Manager.
The practical move is small. Open a test line at the minimum spend, run it on your three highest-intent commercial queries, and instrument it so you can separate the traffic you bought from the traffic you would have received anyway.
That measurement discipline is the entire value of testing early. The learning is cheap right now, and the CPMs are the cheapest they will ever be.
Two hundred days ago this channel did not exist. Your competitors are not faster than you. They are just less attached to the plan they wrote in January.