ChatGPT Ads and B2B: what we genuinely don't know yet

Reach: an unknown, not 15 million

One number is doing the rounds: around 15 million UK adults use ChatGPT in a typical month, measured across web and app in September 2025. There are two problems with treating that as a media plan. It is a consumer usage figure, not an ad-reach figure. And UK ads only serve to logged-in adults on Free and Go, so anyone on Plus, Pro, Business, Enterprise or Education, plus under-18s, is excluded. If someone pays for ChatGPT, or their employer or institution provides it, they cannot see your ad.

So what is the real addressable B2B reach after those exclusions? OpenAI has not said, and I am not going to pretend to estimate it. The honest planning number is: unknown. And keep Ofcom's finding that 54% of UK adults use some AI tool out of this entirely. That covers Gemini, Copilot and the rest, not ChatGPT.

Targeting you can''t verify

The polite framing is "contextual, not audience, targeting." Put less charitably, you cannot target by job title, company size or account. ChatGPT matches on the conversation, and you pay to appear against intent you cannot tie to a person or a company. For a discipline built on knowing exactly who you are reaching, that is not a feature, it is a gap. You may reach the right buyer. You have no way to confirm that you did.

The examples are theory, not evidence

The scenarios people reach for are genuinely useful for one thing: seeing how differently buyers now query, from a quick lookup to a detailed brief. A quick one:

Which CRM is best for a growing business?

An IT lead scoping an implementation goes much deeper:

We''re replacing our HRIS for a 500-person business and I''m running the technical evaluation. For the platforms on our shortlist, compare SSO and SCIM support, API coverage and rate limits, UK and EU data residency, and realistic implementation timelines and integration effort with our payroll and identity provider. What should I be probing on before we commit?

And a decision-maker at the final gate:

About to add this to our risk stack. What''s the strongest case against it?

They show where an ad could plausibly land, and that is worth seeing. But it is all they show. None of them is proof that an ad served, matched, got clicked and drove pipeline. Nobody has that at any scale for B2B yet. Treat the prompts as a hypothesis to test, not a case already made, and be wary of anyone presenting invented prompts as if they were results.

Ads are not the same as organic visibility

Worth separating, because this is where budgets get muddled. Being visible in ChatGPT and buying ChatGPT Ads are two different jobs. If your buyers are on paid plans, they will not see ads at all, so the visibility that reaches them is organic: whether the assistant recommends you in its answer. That is an AEO or GEO question, with a different team, budget and skill set behind it. Do not let an ads pitch quietly turn into an argument for something else.

Who should test, and who should wait

Be blunt about fit, but not by company size. You cannot predict which of your buyers pay for ChatGPT, or get it through work, so you cannot say one segment is reachable and another is not. What you can weigh is your category and your appetite: does the kind of research prompt people actually type plausibly involve you, and can you tolerate spending against reach you cannot verify? If both are yes, there is a narrow, testable case. If you need to know exactly who you reached, this is not it.

And treat "it will get better" with care. The optimistic line is that OpenAI will build stronger targeting and measurement to chase B2B budgets. Probably true. But that is a reason to be wary of funding the beta and de-risking someone else''s roadmap, not a reason to rush in. Better tools later are worth waiting for.

There are no B2B benchmarks, no verified reach, no published case studies, and no way yet to confirm who you reached. That does not make it worthless. It makes it a small, clear-eyed experiment for the few whose category genuinely shows up in these conversations and who have money they will not miss, and a watching brief for everyone else. When the real case studies land, I will change my view as fast as the evidence does. For now, under-claiming is the only honest position.