In early September, trade press reported that OpenAI had told advertising partners it would no longer accept advertising from competitors in generative AI, with stand-alone image and voice generation products named specifically. The change was not publicly announced; it was described as reaching partners directly, and the reporting traces back to The Information rather than to a statement from OpenAI. Treat the detail as unconfirmed.
The structural fact underneath it does not depend on the detail being right, and it is the part worth planning around. Every AI answer surface now accepting advertising is owned by a company that also sells products. Some of those products compete with the advertisers. There is no regulator, contract or norm that obliges a platform to sell inventory to a company it competes with, and platforms have historically declined to.
This is not new. It is newly relevant
Category restrictions are ordinary platform behaviour. Retail media networks have always given their own private label favourable treatment. App stores have restricted ads for competing stores. Marketplaces limit what a competing marketplace may advertise. Nobody considers this scandalous; it is a consequence of buying attention from a company that also sells things.
What is new is that the AI answer layer is being sold to advertisers as the successor to search, and search had a different character. Google's search advertising business has been broadly indifferent to what an advertiser sells, because Google was not selling a competing product in most categories. An assistant is different: it is a product with a roadmap, and that roadmap eats categories — writing tools, image generation, voice, code, customer support, shopping. Every quarter, a category that was neutral becomes adjacent.
Who should care, and how much
If you sell to consumers in a category no AI lab is entering, this is a footnote. If you build software, and particularly if any part of your product does something an assistant does, it is a planning constraint.
- AI-native products — image, voice, video, writing, coding, agents — are the clearest case. Inventory on a rival's surface may be unavailable, and availability today is not a commitment for next quarter.
- Horizontal software with an AI feature is a grey area. The category boundary is drawn by the platform, not by you, and it moves.
- Agencies and consultancies selling AI implementation sit oddly close to the line: the work is complementary, but the positioning language often is not.
- Everyone else should note it and move on — but should still not build a channel plan with a single-source dependency.
The concentration risk is the real exposure
A category restriction hurts most when the restricted channel is the only one that works. Which makes the useful question not "will I be excluded" but "what happens to my pipeline if one surface stops selling to me in thirty days".
That is the same question every advertiser should have been asking about platform concentration anyway, and the answer is the same as it always was: know what share of your acquisition depends on a single owner, and keep at least one route to your audience that nobody else controls. Owned audiences, editorial presence, earned citation in AI answers, partnerships, events. None of them scales the way paid does. All of them survive a policy email.
What to do about it, concretely
Read the policy, not the sales deck
Before planning spend on any AI surface, find the published advertising policies and the restricted-categories list, and keep a dated copy. Platform policies change without announcement, and a dated copy is what turns "they changed the rules" into a documented fact you can act on with a client.
Ask the question in writing during onboarding
If a client's product could plausibly be read as competing with the platform's own, ask the platform or the reseller directly, in writing, before the plan is signed. A verbal reassurance from a sales contact is not a policy, and the person who gave it will not be the person who enforces it.
Write the dependency into the plan
State, in the media plan, what happens if a channel becomes unavailable: which budget moves where, and what the expected cost of that move is. This takes a paragraph, and it converts a crisis into a pre-agreed decision. It is the same argument for writing failure modes into agency contracts rather than discovering them in a quarterly review.
Keep the creative portable
Assets built to one platform's specification with one platform's tooling are harder to move than assets built to a brief. Where a channel carries a policy risk, that is an argument for keeping production independent of the channel that might close.
How to say it to a client
The subject invites alarmism, so the phrasing matters. The wrong version is "the platforms may ban you". The right version is a sentence in a plan: this channel is owned by a company that sells a product adjacent to yours, its policies can change without notice, and here is what we would do with the budget if it did. Said that way it is ordinary risk management, and it is the kind of sentence a client remembers you for when something does change.
The uncomfortable part
There is a version of this story where advertisers complain and a platform relents. It is not the likely version. Selling advertising to a direct competitor while building the same product is a choice no commercial team enjoys making, and the pressure runs one way.
So the planning assumption for the next few years should be that AI answer surfaces are commercially useful, increasingly measurable, and conditionally available — available on terms set by a company whose interests are not aligned with yours in every category. That is not a reason to avoid them. It is a reason to hold them loosely, and to keep building the demand that does not depend on them.
Sources
- https://www.mediapost.com/publications/article/417778/openai-will-no-longer-take-competitive-ads-on-chat.html
