Ads In ChatGPT: How They Work & Whether B2B SaaS Should Use Them

Most B2B SaaS companies should not put budget into ads in ChatGPT yet, and a few should run one small, capped test. The ads are sponsored units below ChatGPT's answers, shown only to free and Go users, with contextual rather than job-title targeting, so they suit cheap-to-buy, self-serve products and rarely suit enterprise deals.

That holds while ChatGPT ads remain a US-first beta with aggregated reporting only. It flips for products whose buyers are individuals on free or Go plans and whose first conversion happens within days.

OpenAI is still changing eligibility, bidding and minimum spend, so check every figure against OpenAI's own pages before you commit budget.

What Ads In ChatGPT Are And How The Rollout Has Gone

Ads in ChatGPT are sponsored units that OpenAI shows below the answer ChatGPT gives, separate from the answer itself. A sponsored unit is a paid placement, so an advertiser buys space beside a response in ChatGPT, the way a search advertiser buys space beside organic results.

That is a different thing from being recommended inside the answer. Paid placement and organic mention are separate channels, and our explainer on what AI search visibility is covers the organic side.

The rollout has moved fast, and every OpenAI-specific fact in this guide carries its source because a beta shifts month to month. The dates that matter are these.

• 9 February 2026: ChatGPT ads began in the US for Free and Go tier users, first through a managed pilot, as MacRumors reported.
• 5 May 2026: OpenAI opened a self-serve Ads Manager in beta to US businesses, which Search Engine Journal covered.
• Later in 2026: OpenAI added more ways to buy, including click-based bidding alongside view-based bidding and a conversions option.

Availability beyond the US is limited and still changing, so a company outside the US should check Ads Manager before planning anything. This guide skips interface walkthroughs on purpose. Setup screens change often inside a beta, and the decisions that cost money sit upstream of any button.

Who Sees Ads In ChatGPT And Who Does Not

Ads in ChatGPT reach free and Go-tier users. Plus, Pro, Team and Enterprise users do not see them.

The size of the paying group matters for B2B. Adriel puts the number of paying subscribers at about 50 million in April 2026, roughly 5% of about 1 billion monthly users. That is a secondhand figure, so read it as a rough size.

Even as a rough size, it tells you where the audience sits. Most of the people using ChatGPT are on free plans, which is where the ads appear. A buyer who opens ChatGPT through a paid work seat never sees your ad.

That is the first filter of the fit test later in this guide. If your buyers do their research on company-paid accounts, the audience you are paying to reach is mostly somewhere else. If the wider question is where your buyers search at all, AI search engines compared lays out the other engines worth checking.

Ad Formats, Placement And Targeting

Placement is simple. The ad is a sponsored unit below the answer, and it stays visibly separate from the answer. Format options beyond that are best read from Ads Manager itself, since creative specifications are the kind of detail that changes during a beta.

Targeting is where B2B teams get surprised. Targeting is contextual, based on topic-level context hints. There is no keyword targeting, no job-title targeting, no company-size targeting and no account-list upload.

In plain terms, you reach a person in the middle of a topic. You do not reach a role. Picture someone asking ChatGPT how to compare project management tools. If your product fits that topic, your ad can appear. Whether that person is a head of operations at a 500-person company or a student writing an essay is invisible to you, and you have no filter to separate them.

Compare that with how most B2B paid programmes are built. A LinkedIn campaign starts from who the person is. A ChatGPT campaign starts from what the person is asking about. Both can work, but they fail in different ways. The firmographic route wastes money on the wrong moment. The contextual route wastes money on the wrong person.

This changes how you write the ad and the landing page. Without a job-title filter, the copy has to qualify the reader. Say who the product is for and what it costs to start, so the wrong visitors leave before they click. Every click is billed, and you cannot screen clicks by title afterwards.

It also changes how you read results. A high click count from a contextual placement tells you the topic is interesting. It tells you nothing about whether the clickers can buy.

Bidding, Budgets And Minimum Spend

There are three ways to buy.

OpenAI added CPC bidding alongside CPM, and under CPC an advertiser is charged only when a user clicks.

The three models are views (CPM), clicks (CPC) and conversions (oCPC, billed per click), as TripleDart's guide lists them.

For a B2B SaaS test, the middle row is the safe starting point. Paying per click caps the cost of a bad audience at the number of clicks you buy. Paying for views offers no such cap on attention that never turns into a visit. The conversions option needs event volume that many B2B offers do not produce, which the measurement section below explains.

Minimum spend is unsettled. OpenAI's announcement does not state a minimum, while TripleDart's guide reports 25 USD per day per campaign. Read Ads Manager before planning, because that is the only place that settles it for your account on the day you launch.

This guide quotes no click or view prices. Take the suggested bid from Ads Manager and use it as the input in the worked example below. A budget built on your own account's numbers beats one built on a number from a blog post, this one included.

Two budget rules apply whatever the bid turns out to be. Set a hard cap before the campaign goes live, and set it at the budget floor you calculate later, not at whatever the account will let you spend. And keep the money separate from your proven channels. A ChatGPT test that quietly eats LinkedIn or search budget makes both results harder to read.

How Conversion Measurement Works

Search Engine Journal's launch report explains that conversions are measured through the OpenAI pixel and the Conversions API, with aggregated reporting only. Aggregated reporting means the platform shows summarised results. It does not give you a list of the people behind the clicks.

Bidding on conversions has its own conditions. Four things must be in place.

• OpenAI pixel or Conversions API: One of the two must be installed and sending events.
• A defined conversion event: The campaign needs one named event to optimise toward, such as a booked demo.
• Advanced Matching: TripleDart's breakdown lists the feature as required for the Conversions objective.
• Consistent event volume: Many B2B events, demo requests included, may be too infrequent for conversion bidding to work.

The last item is the one that catches B2B teams. A high-volume consumer app logs events constantly. A sales-led SaaS company may log a few dozen demo requests a month from every channel combined. Optimisation needs a steady stream of events to learn from, and a trickle gives it nothing.

So plan to judge the test somewhere else. A B2B test should be judged on pipeline, influence and lead quality rather than click-through rate. In practice that means the platform's numbers tell you whether the ad ran, and your own CRM tells you whether it worked.

If you also track organic presence in AI answers, AI search visibility metrics sit next to these paid reporting limits and are worth setting up in the same dashboard. Paid and organic need to be read separately, but a team that can see both is harder to mislead.

ChatGPT Ads Vs LinkedIn Ads For B2B

The deciding difference between ChatGPT Ads and LinkedIn Ads is targeting. LinkedIn offers firmographic targeting by title, company, industry and seniority plus account-list uploads, while ChatGPT ads do not.

The table does not say one is better. It says they answer different questions. LinkedIn lets you choose the person. ChatGPT lets you choose the moment. For a product whose buyers sit in a narrow set of titles at named accounts, the first matters more. For a product bought by individuals who are researching a problem right now, the second may matter more.

One more shift sits behind the comparison. Buyers now start research in places that ad platforms were not built around, which is the subject of our piece on what happened to SEO after AI search. That is a reason to test ChatGPT. It is not a reason to move money before the test reads out.

Judge ChatGPT at about 30 days against your own LinkedIn benchmarks, then wait at least one full sales cycle before moving budget ratios between the two, as Single Grain advises. A 30-day read on a six-month sales cycle tells you about lead quality and little about revenue.

The B2B SaaS Fit Test

Before any budget math, ask whether the people who see the ad are the people who buy. Five questions decide that, and a no on question 1 or 2 ends the test. These are TRM's own working rules, written as judgement and not as sourced data. Move the lines to fit your business.

Question 1, buyer plans. Ads in ChatGPT reach free and Go users and no paid tier. If your customers are founders, freelancers or small-team buyers who use ChatGPT on a personal plan, they are in range. If your customers are managers at larger companies with company-paid seats, they are largely out of range. Fail this one and stop.

Question 2, existing signal. GrowthSpree's guidance says a B2B test suits a considered purchase where ChatGPT already appears in referral or self-reported attribution. Check your analytics for chatgpt.com referrals and read the free-text answers on your demo form. If neither shows ChatGPT, there is no evidence yet that your buyers use it to research, and paying to appear there is a guess. Fail this one and stop too.

To check whether ChatGPT already mentions your product, use ChatGPT answer monitoring tools. A team that needs tooling to run that baseline can start with the AI visibility platform buyer guide. If you would rather see where you stand without a call, the free AI Visibility Report shows where the product sits in AI answers before you pay for ads.

Question 3, purchase type. A mid-funnel ad works on someone who is comparing options. A sub-day impulse buy gives the ad no time to matter, and a multi-year enterprise procurement means the ad reaches a user who is not the buyer. The sweet spot is a purchase that takes days to a few weeks and involves a real comparison.

Question 4, conversion tracking. You need one event, such as a booked demo or a qualified trial, that you can trace from the ad to your CRM. Without it, the only number left is clicks, and clicks are the metric this guide tells you not to judge on.

Question 5, the budget floor. The worked example below turns your numbers into a minimum budget. If that floor is more than you can afford to lose, the answer is no, whatever the other four said.

Pass all five and you have earned a small, capped test. Fail question 1 or 2 and the right move is to spend the time somewhere else.

The Go/No-Go Rule By Deal Size And Sales Cycle

Go, conditional go or no-go depends on average contract value, sales cycle length and buying-committee size, and the cut-offs below are TRM's judgement rather than sourced data.

The first row is where the channel fits today. A buyer on a free or Go plan can see the ad, click, start a trial and decide within days, so a 30-day test produces a real answer.

The second row is where most B2B SaaS companies sit, and it is the row that needs discipline. The test is worth running only if the first two fit-test questions pass. Because the cycle runs one to three months, a 30-day read shows lead quality, not revenue. Judge on qualified conversations at day 30, then check again after a full sales cycle before moving any budget.

The third row is a no. The ad reaches a free or Go user, the pipeline signal arrives long after a 30-day test ends, and the committee that signs is nowhere near the placement. Put the money into organic AI visibility instead.

Your margins may justify different lines. A product with high gross margin and a patient finance team can stretch the cut-offs. A product with thin margins should tighten them. The point of the table is to force the question before spend, not to hand down numbers.

Worked Example: What A 30-Day Test Costs

The budget floor for a 30-day test equals target clicks times your bid. Every input below is an illustrative placeholder, not a benchmark. Replace each one with your own numbers.

1. Pick your bid: Take the suggested bid shown in Ads Manager and call it B.
2. Set the click target: TRM's rule of thumb is about 300 clicks in 30 days, which is enough to learn something and not enough to waste a quarter.
3. Work out the budget floor: Multiply 300 clicks by B.
4. Estimate opportunities: Multiply the clicks by your click-to-lead rate, then by your lead-to-opportunity rate.
5. Find the allowable cost per opportunity: Multiply ACV by win rate by the share of first-year revenue you will spend on acquisition.
6. Compare the two costs: Divide the budget floor by expected opportunities to get cost per opportunity. If that is above the allowable figure, the test fails on paper.
7. Check your loss limit: If the budget floor already exceeds what you can afford to lose, stop here.

Here is one run with hypothetical numbers.

On these placeholder numbers the test fails, because 1,250 USD per opportunity is above the 900 USD the deal can carry. That is a useful result. It shows which input has to move. A lower bid, a better click-to-lead rate from a sharper landing page, or a larger deal would each change the answer, and you can test each on paper before spending anything.

The placeholders are not predictions. A 2% click-to-lead rate and a 20% lead-to-opportunity rate are round numbers chosen to make the arithmetic readable. Your own funnel data replaces them. If you have no click-to-lead data from comparable paid traffic, say so in your planning notes and widen the margin, because the weakest input drives the result.

A 30-Day Test Design With Kill Criteria

If the fit test and the worked example both pass, run the test in this order. The rules here are TRM's working rules.

1. Set the single conversion event: Use a booked demo or a qualified trial, not a click.
2. Install and verify tracking: Put the OpenAI pixel or Conversions API in place and confirm the event fires before spending.
3. Start on click bidding: Use conversion bidding only when event volume supports it.
4. Cap total spend: Set the cap at the budget floor from the worked example.
5. Hold one landing page and one offer: Changing the bid and the page in the same week makes the result unreadable.
6. Review weekly and decide at day 30: Weekly reviews catch broken tracking. Decisions wait for the full period.

Extend only when the deal cycle is longer than 30 days, and then only to a second checkpoint after one full sales cycle. The first read shows lead quality. The second shows whether those leads closed.

On a pass, run one more capped cycle with the same event, the same cap and one change at a time. Never scale up immediately. A single good month on a channel still in beta is a reason to repeat the test, not a reason to rebuild the budget around it.

If The Answer Is No: Earn Mentions Without Paying

When the fit test fails, the budget does better somewhere else. Organic mentions in ChatGPT answers are earned through content and citations, and they are measured separately from ad spend. That makes them a different channel with a different timeline, and a failed ads test does not close it.

The practical starting point is the baseline from the fit test. Find out which questions ChatGPT answers about your category and whether your product appears in them. From there, the work is content that answers those questions clearly and earns citations on pages ChatGPT draws from.

These reads cover the route in order.

• Roadmap: The AI search visibility strategy post sets out where the budget that is not going to ads should go.
• Playbook: The AEO strategy for B2B SaaS post covers the organic route in detail.
• Measurement: Use tools for tracking brand visibility in AI search to measure whether it is working.

If you would rather hand the work to a specialist, TRM's generative engine optimization work is the service built around it.

Common Mistakes When Testing Ads In ChatGPT

• Judging on CTR instead of pipeline. Click-through rate says the ad interested someone. Judge on qualified conversations and lead quality in your CRM.
• Testing an enterprise offer on a free-tier audience. The ad reaches free and Go users, not procurement committees. Keep enterprise offers off this channel and send the budget to organic AI visibility.
• Starting conversion bidding with too few events. Conversion bidding needs consistent event volume, and demo requests may be too infrequent. Start on click bidding and move only when volume supports it.
• Changing the bid and the page in the same week. Two changes at once make the result unreadable. Change one thing, wait a week and record what moved.
• Reading a 30-day result as the final answer for a long sales cycle. Thirty days shows lead quality and little else on a long cycle. Wait for a full sales cycle before moving budget ratios.

If the test fails and the money needs a home, answer engine optimization is the route aimed at earning citations in answers without paying per click.