Updated 25 October 2026.
If your buyer is a director paying for their own ChatGPT subscription, they will never see your ad. It is the first thing to check before you spend a euro on ChatGPT Ads in B2B, and the one most often skipped, even though a single question about your buyer settles most of it. Part of the B2B market has no business on this channel today; this article is here to help you work out whether you belong to it.
Who sees ads in ChatGPT?
Signed-in adult users on the Free and Go plans. Nobody else.
Plus, Pro, Business, Enterprise and Edu subscribers see no ads at all. OpenAI documents this, and it is a delivery rule rather than a campaign setting: no budget, no bid and no context hint lifts it.
For B2B the consequence is immediate. The more someone uses ChatGPT for work, the more likely they are to sit on a paid plan, or on a Business or Enterprise licence their employer rolled out. So the channel reaches people who have the problem first, and only sometimes the ones who hold the budget.
The B2B buyers you can still reach
Operational buyers, SMEs, founders and the roles that test tools. The buying committee of a large group stays out.
| Buyer profile | Reachable? | Why |
|---|---|---|
| Founder or owner of a small business | Often yes | Personal and work use blur together, Free or Go is common, decides alone |
| Operational buyer (HR manager, office manager, sales manager, project lead) | Yes | Describes the problem to ChatGPT before knowing the solution category, often on a personal account |
| Roles that test tools (growth, ops, product, marketing) | Yes, with a caveat | Strong appetite for new tools, but as soon as the company rolls out Business or Enterprise they leave the inventory |
| Senior decision-maker in a large group (CIO, CFO, VP) | Rarely | Likely on a work subscription, company account, committee purchase |
| Target under a deployed Enterprise or Edu licence | No | The account receives no ads |
| Public-sector buyer or tender cycle | No | A decision that takes more than nine months cannot be learned with a test budget |
The table carries no audience share, for a good reason: nobody has one. OpenAI publishes neither the split between plans nor any benchmark by industry, so anyone quoting you a reachability percentage made it up.
How do you estimate reach before signing?
With three questions about your buyer, which lead to a decision rather than a percentage.
The first is about profile: who, inside your client, opens ChatGPT to describe the problem you solve? If it is the person who signs, the channel is direct. If it is someone who then has to persuade their director, the channel feeds the top of the journey, and the landing page must hand them the arguments.
Company size weighs heavily next. The bigger the company, the likelier it is that a Business or Enterprise licence went out to every employee, and then the buyer drops out of the inventory even with the right profile. An SME with no formal AI policy, by contrast, leaves its teams on personal accounts.
Last comes the likely use of a paid plan. Heavy daily use, confidential data or contractual confidentiality requirements push toward paid; occasional use, to understand a topic or compare options, often stays on Free or Go.
After that, the only real test is the test. Run a campaign with a learning goal rather than a scaling one, starting from the minimum budget of $25 per day per campaign. Above all, build the full measurement chain on day one: UTMs, the OpenAI Pixel, the Conversions API, and the CRM to read lead quality, not just lead count. The native report stops at six columns. The answer to “are my buyers there?” shows up in the CRM, a few weeks in. Not before.
Our three-branch qualification
Before any commitment we place each advertiser in one of three branches, and B2B lands in the middle one: yes, with a caveat.
Yes. B2C or B2B2C, average basket above €100, CAC up more than 30% over eighteen months. The audience is there and the volume allows learning.
Yes, with a caveat. B2B. Ads only reach the Free and Go plans, so a senior decision-maker is often out of reach; the deal stays workable when the buyer is operational, meaning they look like the first three rows of the table. We say so before signing, not after the first report.
No. Monthly media below €3,000, or a sales cycle longer than nine months. Volumes do not allow learning, and the channel is not built to carry a long decision. We say that too, and we give the reasons. No hedging.
The grid is the same on our home page and on a call, and it can perfectly well end in a refusal.
Does operational B2B actually work?
Yes, on at least one documented case: a French HR SaaS company, a B2B target in active comparison, with a sales cycle built around qualified demos.
The typical buyer is an HR manager or SME owner describing a recruiting or applicant-management problem. They match the second row of the table. For them, a context hint describes a situation rather than keywords: “HR team spending too much time chasing candidates by email”, not “hr software, ats, recruiting”.
Results published in the Leverage white paper of September 2026: more than 30 qualified demos per month, 12 SQLs, a CPL of around €250, a demo-to-SQL conversion rate 70% above the Google Ads benchmark, and a final customer CAC 22% below Google Ads.
Do not read this case as a promise of generalisable results. It establishes one precise thing: when the buyer is operational and describes their need to ChatGPT, the channel produces a usable signal within the first weeks. It says nothing about a senior buyer under an Enterprise licence. The details are in the B2B lead generation use case, and the comparison with Search is in ChatGPT Ads vs Google Ads.
If your buyer is not reachable, we will tell you before any commitment, even if it means turning the work down.