How to Decide Whether ChatGPT Ads Can Reach Your B2B Buyers Using Your Last 5 Closed Deals
ChatGPT ads run on the Free and Go plans only, so every buyer on a seat their company pays for is unreachable at any budget. Run the 3 checks in this article against your last 5 closed won deals and you will know in an afternoon whether to spend or skip.
Will Leatherman
Founder, Catalyst

TLDR
ChatGPT ads appear on the Free and Go plans only, and Plus, Pro, Business, Enterprise and Education seats never see one. Before you commit budget, run 3 checks. Who can see the ad, what the public numbers actually show, and whether your own buyers sit inside that audience. Pull your last 5 closed won deals and ask whether each company pays for ChatGPT and whether they could have started without a meeting. If 0 or 1 of the 5 could have seen an ad, skip the channel. If 2 or more could, test at the $25 a day minimum for 2 to 4 weeks and report cost per company that fits instead of cost per click.
A ChatGPT ad is one sponsored card placed below the assistant's answer, sold on impressions or clicks, and it reaches people using the free tiers of ChatGPT. That last part settles most B2B budget questions on its own. OpenAI's help center says ads show on the Free and Go plans only, and the paid tiers above those never see any ad at all. A buyer researching vendors from the Business seat their employer pays for is outside the audience no matter what you bid.
The pull toward the channel is real. G2 surveyed over 1,000 B2B software buyers this year and about half now start research in an AI chatbot rather than Google. Graphite and Similar Web tracked US ChatGPT ads from March to August and found B2B software was the single biggest category, roughly 1 in 5 of every ad shown, with monday.com leading as the first adopter. The buyers moved and the ad money followed.
"In the next 30 minutes, you're going to learn three checks that tell you whether ChatGPT ads can actually reach your buyers using your last five closed deals." — Will Leatherman, Founder of Catalyst
Open your last 5 closed won deals before you read on. You will mark them in check 3.
What are the 3 checks to run before you buy ChatGPT ads?
Run them in order, and each one can stop the spend on its own.
| Check | The question | What it rules out |
|---|---|---|
| 1. Who sees the ad | Which ChatGPT seats can an ad reach at all | Paying to reach buyers whose plan hides every ad |
| 2. What the first numbers show | What have real advertisers spent and received | Planning off a forecast nobody has matched in practice |
| 3. Are your buyers in it | Do your own last 5 closed deals sit in that audience | A budget set before anyone checked the audience |
Every one of these starts in your own CRM rather than in a platform dashboard. That matters because the platform is also the scorekeeper. Gartner predicts that by 2028, 80% of US ad spend will run through platforms where AI sets the buying, the cost and the outcome, and its advice to CMOs is to get independent proof. One agency told Digiday that ChatGPT reported about 100 clicks where its own tracking found only 20.
"The platform reports its own results and somebody has to check them." — Will Leatherman, Founder of Catalyst
Who actually sees a ChatGPT ad?
People on the Free and Go plans, and nobody else.
The ad is a labeled card in a separate section under the response. Most teams picture a sponsored result sitting at the top of the answer the way it works on Google. ChatGPT keeps the two apart. OpenAI shows 1 ad unit below the response marked sponsored, often below the fold, and advertisers cannot shape, rank or change the answer itself. Ads have been in testing since February 9th.
| ChatGPT plan | Sees ads |
|---|---|
| Free | Yes |
| Go | Yes |
| Plus | No |
| Pro | No |
| Business | No |
| Enterprise | No |
| Education | No |
One buyer can sit on both sides of that table in the same week. They ask from a free account on their phone and from the enterprise seat their company pays for, and only one of those sessions can carry an ad. Nobody has measured which seat B2B buyers use when they research vendors. A single LinkedIn poll tried and the sample was too thin to rely on.
"Who sees it depends on the seat and nobody has counted the seats." — Will Leatherman, Founder of Catalyst
That gap is exactly why check 3 uses your deals instead of someone else's benchmark.
What do the first numbers from real B2B accounts show?
Thin volume, high cost per click, and results that arrive one account at a time.
You can buy 3 ways, CPM per thousand views, CPC per click, and oCPC, which still bills on the click. The floor is $25 a day.
"Nothing bills per conversion. So this is purely a top of funnel play." — Will Leatherman, Founder of Catalyst
| Advertiser or study | What they spent | What came back |
|---|---|---|
| Symphonic Digital | $3 a click against a forecast the platform called strong | 0 impressions served, real clicks landed near $9 |
| Workshop Digital | $2,319 of its own money | 739 clicks and 0 qualified leads |
| A B2B advertiser on a LinkedIn lookalike | $7,000 CAD | 336 clicks across 146 companies, 5 of which fit, 0 deals |
| Graphite and Similar Web US panel | Panel of US ChatGPT ad exchanges | 0.073% click rate, well under 1 in 100 |
| Intelligencia, financial services account | Not disclosed | Started at 3.2x paid search cost per approved lead, 35% below search by month 6 |
The $7,000 CAD test is the one worth holding onto. Clicks matched to 146 companies and 5 of them fit the profile, which works out to about $1,400 CAD per company that fit, with no deal attached. The other 141 are the whole problem, and check 3 explains them.
Counter evidence exists and it carries a caveat. One advertiser site with roughly 1,500 visitors from each source saw Google Ads convert at 3.71% and ChatGPT ads convert at 21%. That figure comes from a Search Engine Land writer who sells paid search. Another agency reports SaaS clients landing under half their Google cost per lead without specifying the spend behind it.
"OpenAI says it doesn't have performance benchmark metrics yet, which to me is not a great sign." — Will Leatherman, Founder of Catalyst
"Every result so far, good or bad, is just one specific account." — Will Leatherman, Founder of Catalyst
Who the clicks land on explains most of the spread. We read every pricing page of Graphite's 10 largest B2B advertisers and 8 of the 10 let a visitor start free or start a trial without talking to sales, including monday.com, HubSpot, Zapier and Semrush. Pylon sells by demo only. For the self serve 8, a click can become a signup the same day. If you sell through a sales team, that same click still has to become a meeting, and nothing in the auction prices that for you.
When does ChatGPT actually show an ad?
Most often when somebody has already named a brand.
Graphite found a question about Wix drawing an ad for Lovable, and in 61% of the exchanges that carried an ad, a brand had been named in the conversation.
"Your competitor's name is the trigger." — Will Leatherman, Founder of Catalyst
That makes the ad an answer to a conversation already running about somebody else, which is the same dynamic covered in why ChatGPT names your competitor instead of you.
The largest advertisers spend like they believe that. Digiday describes an agency client spending $10 million a month on Google and putting under $100,000 a month into ChatGPT. Graphite watched Zapier and SAP increase their spend and then stop entirely.
Does being cited in the answer change what the ads do?
Yes, and the lift shows up on the ads you already run.
SE Ranking ran over 50,000 US commercial prompts and found the advertiser was cited in the answer above its own ad about 4% of the time. The same work on Google found brands cited in the AI Overview took 15.7% paid click-through against 11.2% for brands that were not cited.
Every seat sees the answer. Only 2 plans see the ad. That asymmetry is why earning the citation outranks buying the card for most B2B teams, and it is the work laid out in 3 plays that turn AI search into a revenue channel.
Are your buyers actually in the audience?
Answer it with your last 5 closed won deals rather than with a benchmark.
Pull them from your CRM and ask 2 questions of each one.
- Does that company pay for ChatGPT? A Business or Enterprise rollout means the buyer never saw an ad.
- Could they have started without a meeting? Self serve buyers can sign up from a click. Sales led buyers cannot.
If you cannot answer either question, your champion can. Ask on the next call what they use ChatGPT for, whether they search in it, whether it records their calls. Much of this is already sitting in your call recordings.
| How many of your 5 could have seen an ad | What to do |
|---|---|
| 0 or 1 | Skip the channel |
| 2 or more | Run a bounded test |
"The rule on this slide is our reading of the evidence, not a study." — Will Leatherman, Founder of Catalyst
The reasoning behind it is short. Paid seats never see the ad, so clicks lean toward people on Free or Go wherever they happen to work. Sales led companies need a meeting, and a click does not deliver one. That is the most likely explanation for the 141 companies that did not fit.
What does a defensible test look like?
$25 a day, 2 to 4 weeks, on prompts that name you or a competitor, reported on companies rather than clicks.
Point the budget at prompts where a brand gets named, because that is where ads fire. Give it the full 4 weeks before reading anything into it. Then count the companies that fit per 100 clicks.
"Report cost per company that fits, never cost per click. That's the number your boss will ask about." — Will Leatherman, Founder of Catalyst
The $7,000 CAD test above produces $1,400 CAD per company that fit and 0 deals. Cost per click would have made the same test look fine. The same discipline applied to measurement generally is in 3 checks that tell you whether your AI visibility score really changed, and the conversion benchmarks to compare against are in B2B conversion rates from AEO.
Where does the attention go if the clicks never arrive?
Into the chat window, which is why click volume keeps falling while buying continues.
Bot traffic has been inflating paid click counts on Google for years, and an AI visibility tool can separate bot visits from human ones on your own site. The larger shift sits underneath that. Buyers now read what they need about a vendor without opening the vendor's site.
"You can now learn everything you need to learn about a company without ever leaving your ChatGPT session." — Will Leatherman, Founder of Catalyst
Traffic falls and purchase decisions keep happening, which shows up as a rising conversion rate on the visits that remain. Measuring the answer rather than the session is the practical response, and the walkthrough is in how to audit your AI search ranking in 20 minutes.
The takeaway
Half of B2B software buyers now start in a chatbot, and the seats their companies pay for will never show them an ad. Your last 5 closed deals settle the question faster than any benchmark will.
"When someone asks, 'Are we on ChatGPT ads?' your answer is your last five deals." — Will Leatherman, Founder of Catalyst
Mark those 5 deals today and ask 2 champions this week which ChatGPT plan they actually use. If 2 or more could have seen an ad, run $25 a day for 4 weeks and report cost per company that fits. Whatever seat your buyer sits on, they see the answer, so start there with the free AI visibility audit, which asks 5 AI tools your real buyer questions and returns a report in about 5 minutes. The prompts behind that work are in the AEO skills pack.
The Content Engineer
Enjoyed this article?
Frameworks like this, weekly. No fluff, just original research and actionable insight.
More in AEO

How AlphaSense Wins 72.9% of AI Voice Share and the 4 Fixes You Can Copy This Week
AlphaSense holds most of the AI answers in market intelligence, and the thing carrying it is not budget. It is 41 pages that each answer one question a buyer actually types. Here is the teardown of 6 of their pages and the 4 fixes you can run on your own site, cheapest first.

3 Checks That Tell You Whether Your AI Visibility Score Really Changed
A B2B brand went from 53% to 86% share of voice in 6 weeks with nothing changed on its website. Run the 3 checks in this article against your own last reported number and you will know within an hour whether the move was real or whether your tracker simply asked different questions.

How Ramp Wins 40% of AI Search Share of Voice With a 29 out of 100 Visibility Score
Ramp gets named first when a finance lead asks an AI engine for the best spend management software, and it scores in the bottom third on an AI visibility audit. The gap between those two facts is the whole lesson, and 5 of the 6 fixes it points to cost you an afternoon.
Ready to turn insight into pipeline?
We work with B2B companies that know content is the moat. Let's talk.