OpenAI’s advertising business has reached a $1 billion annualised revenue run rate in under 200 days, with tens of thousands of advertisers, according to OpenAI’s own announcement. Ads are now live in over 40 countries, with self-service access opening across India, Europe, the Middle East and North Africa.
For anyone who has spent the last two years building organic visibility inside AI assistants, this raises an obvious and slightly uncomfortable question: what happens to the organic slot now that there is a paid one next to it?
We are going to answer that carefully, because the honest answer includes a substantial “nobody knows yet”, and because we have a stake in it. Our own go-to-market is deliberately organic-first. A development that complicates that position deserves examining rather than dismissing.
What OpenAI has actually committed to
Start with what is on the record, rather than what people are inferring.
On separation and influence, OpenAI states:
“Ads in ChatGPT are always clearly labeled and separate from ChatGPT’s answers. Advertising does not influence the answers ChatGPT provides.”
That is unambiguous, and it is a meaningful commitment. It is not hedged. We take it at face value.
On targeting, OpenAI describes a contextual system: “the ads system uses the context of the current conversation to show an ad relevant to what someone is exploring.” It adds that “depending on the country and a user’s settings, the ads system may also use context from the user’s broader ChatGPT experience.”
On advertiser access to data, OpenAI states: “Advertisers do not receive access to people’s private conversations.”
On who sees ads, the announcement describes an advertising-supported free tier alongside paid consumer subscriptions, enterprise offerings and usage-based APIs.
Taken together, this is a more conservative design than it could have been. Answers and ads are separated. Ad spend is stated not to move the recommendation. That is a better starting position than the one search advertising began from.
What the commitment does not cover
Being precise about scope is not the same as being cynical.
The commitment is that advertising does not influence the answers. That is a statement about the assistant’s output.
It is not a statement about attention — about how a user’s eye and click divide between an answer and a labelled unit sitting beside it. That is a question about human behaviour, not model behaviour, and OpenAI has not claimed anything about it because it is not really theirs to claim.
Nor is it a statement about future interface design. Ad placement, prominence and density are product decisions that evolve. The commitment as written concerns influence on answers, not the amount of screen real estate ads occupy in a year’s time.
This is not a criticism of OpenAI. It is a note that a strong commitment on one axis is not a commitment on every axis, and the difference matters if you are planning a two-year content programme.
What we genuinely do not know
Here is the part where you will see a lot of confident commentary over the next quarter, almost none of it supported.
We do not know how attention divides. In traditional search we have twenty years of data on how users distribute clicks between paid and organic. In conversational AI, we have essentially none, and the interface is different enough that transferring those assumptions would be unsound.
We do not know the B2B picture. Most early reporting concerns consumer and shopping journeys. A finance director evaluating an infrastructure supplier behaves nothing like someone asking for trainer recommendations, and B2B categories may see very different ad density and relevance.
We do not know whether this changes citation economics. If advertising becomes the primary revenue mechanism for AI assistants, does that alter how sources are selected, ranked or displayed over time? There is no evidence that it will. There is also no evidence that it will not. Anyone telling you either way is speculating.
We do not know how this interacts with measurement. Google now reports AI impressions in Search Console, which gives some visibility on Google surfaces. There is no comparable first-party reporting for ChatGPT, so the surface where advertising is scaling fastest remains the one you can measure least.
What does not change
Against all that uncertainty, several things are stable enough to act on.
Eligibility is still the floor. An assistant cannot cite content it cannot reach. That was true before advertising existed and is unaffected by it. If your crawler access is broken — and after 15 September more configurations will be in flux than usual, as we covered in Cloudflare’s AI crawler defaults — no amount of budget fixes it, because the ad unit and the answer are separate systems.
Being the source is a different position from being the advert. When an assistant builds an answer from your material, you are part of the reasoning. When you buy the unit beside it, you are adjacent to the reasoning. Those are different commercial positions with different trust characteristics, and the second does not replace the first.
The underlying work is the same work. Structured, accurate, well-sourced, technically reachable content is what makes you eligible for citation. It is also, incidentally, what makes a landing page worth sending paid traffic to. There is no version of this where that investment is wasted.
Organic is still the only compounding asset. Paid placement stops the day the budget stops. A well-cited body of technical content does not.
What we would actually advise
For a B2B business in September 2026, in order:
1. Fix eligibility first. It is free, it compounds, and it is a prerequisite whichever way the paid question resolves. Most businesses we audit have unresolved technical visibility problems that no budget would fix.
2. Baseline your measurement now. The surface is changing underneath everyone. If you want to know in six months what the arrival of ads did to your organic presence, you need a reading from before it scaled.
3. Treat paid AI placement as an experiment, not a strategy. Self-service is new in most regions. There are no reliable B2B benchmarks. A controlled test with a defined budget and a clear read-out is reasonable; restructuring your programme around it is not.
4. Do not let anyone sell you a forecast. The most reliable indicator of a weak advisor in the next two quarters will be confident numbers about how ChatGPT advertising affects organic citation rates. That data does not exist yet.
Our own position, stated plainly
We run an organic-first programme and we recommend one. This development does not change that, and we want to be clear about why — because “the thing we already sell is still the right answer” is exactly the reasoning you should be suspicious of.
The reason is not that advertising will not work. It may work very well. The reason is that eligibility, structure and citability are prerequisites regardless, they cost nothing but discipline, and they are where almost every business we look at still has unresolved problems. Recommending paid experimentation to a company that is not yet reachable by the assistants would be selling the second thing before the first.
If that calculus changes when real B2B data emerges, we will say so here.
The short version
ChatGPT advertising is real, large and growing fast. OpenAI has committed on the record that ads are labelled, separated from answers, and do not influence what the assistant says.
What nobody can yet tell you is how attention divides, how it behaves in B2B, or what it does to citation economics over time.
Fix eligibility, baseline your measurement, test paid deliberately if at all, and treat anyone offering confident numbers with appropriate suspicion.