Direct answer: When OpenAI Ads (the ad units inside ChatGPT) launched in February 2026, early reports pegged the cost at roughly a $60 CPM with steep entry minimums in the $200,000–$250,000 range, putting it out of reach for most B2B teams. That has compressed fast: OpenAI opened a self-serve Ads Manager to U.S. advertisers and dropped the $50,000 minimum, CPC bidding arrived with observed bids reported around $3–$5 per click, and some later coverage suggested CPMs had softened well below the launch rate. What you actually pay today depends on whether you buy on CPM or CPC, and conversion (CPA) bidding is still on the roadmap, not live, which is the single most important fact for budgeting a long B2B sales cycle.
Key takeaways
- OpenAI rolled ads out inside ChatGPT in February 2026, and early launch coverage described a premium CPM (around $60) with six-figure entry minimums (TechCrunch).
- The economics opened up quickly: OpenAI brought a self-serve Ads Manager to U.S. advertisers and dropped the $50,000 minimum, while promising third-party measurement and CPA bidding for the future (Digiday).
- CPC bidding is now available, with reported bids landing between roughly $3 and $5 per click (Search Engine Journal).
- Cost-per-action style buying is being discussed and tested, but conversion (CPA) bidding is described as roadmap, not a live self-serve buying option you can rely on yet (Digiday).
- Who actually sees ads, and what's allowed, is governed by OpenAI's own documentation and ad policy, which matters because rejected creative wastes budget (OpenAI Help, OpenAI ad policies).
- For B2B, the honest planning number isn't a CPM or a CPC, it's cost-per-qualified-pipeline, and because CPA bidding isn't live, you carry the conversion risk yourself and must budget for a learning period rather than a guaranteed cost-per-lead.
What OpenAI Ads cost at launch: ~$60 CPM and six-figure minimums
OpenAI began rolling ads into ChatGPT in February 2026. The first wave was clearly built for large brand advertisers, not self-serve performance marketers. Early launch reporting described pricing in the neighborhood of a $60 CPM, that is, roughly $60 to show your ad a thousand times, paired with entry minimums that, across the rollout and Ads-Manager coverage, were described in the $200,000–$250,000 range (TechCrunch; Digiday).
A $60 CPM is high relative to most programmatic display, but it is not absurd for premium, high-attention placements, and ChatGPT is about as high-attention as a surface gets. The bigger barrier was the minimum spend. A $200k–$250k commitment is a managed-buy number: it implies an insertion order, an account team, and a brand budget. For a B2B SaaS company spending $5k–$50k a month across channels, that launch configuration was simply not a self-serve option. It was a signal that the inventory existed and was being priced like premium media, not a place to test a campaign next week.
It's worth being precise about what these figures are. They come from early launch reporting, not from a public OpenAI rate card. Treat them as the launch-era benchmark, the starting point of the curve, rather than today's price. The whole point of the next section is that the curve moved.
The compression: self-serve opens, the $50k minimum drops, CPMs soften
Within months, the entry economics changed materially. OpenAI opened up a self-serve Ads Manager to U.S. advertisers and dropped the $50,000 minimum that had stood between smaller buyers and the inventory, while publicly promising third-party measurement and CPA bidding as future capabilities (Digiday). That is the difference between "premium managed buy only" and "a real performance channel a mid-market B2B team can actually test."
Two things happened at once. First, the floor came down, you no longer needed a five-figure minimum just to get in the door through the self-serve manager. Second, some later reports suggested CPMs had softened well below the launch rate as more inventory came online and more advertisers competed in the auction. We're being deliberately careful with that second point: the launch-era $60 CPM is the figure we'll anchor to, and we won't put a hard number on the softened rate because it isn't something we'd attribute to a primary source. The direction of travel, down, is well supported; the exact landing spot varies by category, format, and auction pressure.
The practical takeaway for a B2B buyer is that "OpenAI Ads is too expensive to try" stopped being true sometime after launch. The channel went from an enterprise insertion order to something you can stand up with a normal monthly test budget. That doesn't make it cheap, high-attention inventory rarely is, but it makes it accessible, and accessible is what matters when you're deciding whether to run an experiment this quarter.
CPC bidding arrives: observed bids around $3–$5 per click
The most important pricing development for performance-minded buyers is that OpenAI Ads now supports CPC bidding. Reporting on the rollout described observed bids landing between roughly $3 and $5 per click (Search Engine Journal). That's a familiar shape for anyone who has run search or paid social: instead of paying for impressions and hoping for engagement, you pay when someone actually clicks through.
A $3–$5 CPC is broadly competitive with, and in many B2B categories cheaper than, branded and competitor keywords on traditional search, where high-intent B2B terms routinely run well into the teens or higher per click. The catch is intent quality. A click from a ChatGPT conversation is a different animal than a click on a search ad: the user is mid-task, mid-question, and the ad is appearing in a conversational context. Early CPCs that look attractive on paper still need to be judged on what those clicks do once they hit your site, because a cheap click that doesn't convert is not a cheap lead.
For budgeting, CPC bidding is genuinely useful because it converts an abstract CPM into a unit you can model. If you can hold a $4 average CPC and your landing experience converts visitors to a meaningful next step at, say, 3–5%, you can back into a cost-per-conversion estimate and a daily budget. That's a real plan, as long as you remember it's your estimate, built on your funnel, not a number the platform is optimizing toward for you. Which brings us to the gap.
What's NOT available yet: CPA bidding and third-party measurement
Here is the single most important thing to understand before you set a budget: conversion (CPA) bidding is not a live, self-serve buying mode you can lean on. OpenAI has publicly framed third-party measurement and CPA bidding as things it is promising and testing, not capabilities you can switch on today and trust to optimize toward your cost-per-acquisition target (Digiday). Cost-per-action ad formats are being explored and turned on in early forms (Digiday), but that is not the same as a mature, account-wide CPA bidding strategy of the kind Google and Meta offer.
Why does this matter so much for B2B specifically? On a long B2B sales cycle, the conversion you care about, a qualified demo, an opportunity, closed-won revenue, happens weeks or months after the click, often offline, inside your CRM. CPA bidding on mature platforms works because the platform ingests that downstream signal and optimizes the auction toward it. Until OpenAI's measurement and CPA capabilities mature, the algorithm can't optimize toward your pipeline because it can't see your pipeline. You are buying impressions or clicks, and you are the one responsible for turning those into qualified pipeline.
The honest implication: budget OpenAI Ads in 2026 the way you'd budget an early-stage channel test, not a tuned acquisition engine. Expect a learning period. Expect to do your own conversion attribution. Don't promise your CFO a cost-per-lead the platform is contractually optimizing toward, because it isn't, not yet. When CPA bidding and third-party measurement do land, the economics get a lot friendlier; planning around them now would be planning around a roadmap.
Skip the manual work
Abmatic AI runs targets, sequences, ads, meetings, and attribution autonomously. One platform replaces 9 tools.
See the demo →The OpenAI Ads cost model at a glance
Here's how the three buying models compare today, including the one that isn't live. Use it to decide what you can actually plan around.
| Buying model | Live in 2026? | Typical figure | You pay when | Best for |
|---|---|---|---|---|
| CPM (cost per 1,000 impressions) | Yes | ~$60 at launch; reported to have softened since | Your ad is shown | Brand reach, awareness, category education |
| CPC (cost per click) | Yes | ~$3–$5 observed per click | Someone clicks through | Demand capture, traffic, modeling a cost-per-conversion yourself |
| CPA (cost per action / conversion) | Not as mature self-serve bidding, on the roadmap; early CPA formats emerging | No reliable benchmark yet | (Future) a defined conversion occurs | (Future) efficient acquisition once measurement matures |
Figures: CPM and minimums from launch-era reporting (TechCrunch, Digiday); CPC range from Search Engine Journal; CPA status from Digiday. For a fuller orientation, see our OpenAI Ads 101 guide and ChatGPT Ads explained.
Setting a B2B starting budget (a framework, not a guarantee)
Because CPA bidding isn't live, the right way to set a budget is to size a structured test that produces enough data to make a real decision, not to back into a cost-per-lead and pretend the platform will hit it. Here's the framework we use.
1. Decide what you're buying first. If you want to learn whether ChatGPT audiences engage with your category at all, start on CPM and watch attention and click-through. If you already know the audience and want traffic you can convert, start on CPC, where a ~$3–$5 click gives you a unit to model.
2. Size the test off your funnel, not a target CPL. Take your CPC (use $4 as a working assumption), your realistic landing-page conversion rate to a meaningful next step (often 2–5% for cold B2B traffic), and the volume of conversions you need before you'd trust the result (50–100 next-step conversions is a reasonable read). At $4 CPC and a 3% conversion rate, each conversion costs about $133 in media, so 75 conversions is roughly a $10,000 test. That's a planning number you can defend, not a promise.
3. Budget for a learning period. Set aside the first portion of spend (think 20–30%) as explicitly exploratory, creative variants, audience signals, and landing pages you expect to throw away. Treat that as the cost of buying information in a channel where the algorithm can't yet optimize toward your pipeline.
4. Instrument the downstream conversion yourself. Since OpenAI's measurement is still maturing, you have to connect the click to the CRM outcome on your side. Tag every entry, attribute the demo or opportunity it produced, and judge the channel on cost-per-qualified-pipeline, not CPM or CPC. This is exactly the gap that bites teams: a great CPC can hide a terrible cost-per-opportunity, and you won't know unless you wired it up.
If you want this run as a controlled experiment rather than a guess, proper attribution, creative variants, and a clean read on cost-per-pipeline, that's the kind of work we do for customers. Abmatic AI runs OpenAI Ads on the live Ads API alongside your other channels, replaces 6sense, Demandbase, Mutiny, and Qualified, and pipes results into Salesforce, HubSpot, or Marketo; if that's the setup you want, Book a demo.
How Abmatic AI runs and optimizes OpenAI Ads spend
We built a live integration to the OpenAI Ads API and run these campaigns ourselves, so the budgeting framework above isn't theoretical, it's how we operate. A few things we've learned spending real money on this inventory.
We start narrow and let the auction teach us. Because there's no CPA bidding to lean on, we don't dump a large budget into a broad campaign and hope. We run tight audience signals and a handful of creative variants on CPC, read which clicks actually convert downstream, and concentrate spend on the combinations that produce pipeline. The early dollars buy information; the later dollars buy outcomes.
We treat policy as a budget lever. Rejected creative is wasted spend and wasted learning time. We write to OpenAI's ad policy from the start, company- and account-level framing rather than language that trips privacy or competitor-disparagement reviews, so campaigns go live and stay live (OpenAI ad policies; how ads work in ChatGPT). For the full operational walkthrough, see how to run ChatGPT Ads.
We close the measurement loop the platform can't. Every click is attributed to a CRM outcome on our side, so we judge OpenAI Ads on cost-per-qualified-pipeline and compare it honestly against other channels, including search, which we break down in our OpenAI Ads vs. Google Ads for B2B SaaS comparison. When CPA bidding and third-party measurement mature, that loop is already built; we just hand the signal to the platform.
The net of it: OpenAI Ads in 2026 is a real, accessible B2B channel with a CPC you can model, a CPM you can use for reach, and a conversion-bidding gap you have to cover yourself. Run it as a measured experiment, instrument the pipeline, and you can get a clean answer on whether it earns a permanent line in your budget. If you'd rather have a team that already runs it on the live API handle the experiment end to end, Book a demo.
Keep reading
- OpenAI Ads 101: the start-here guide
- OpenAI Ads vs Google Ads for B2B SaaS
- How to run ChatGPT Ads (step-by-step setup)
- ChatGPT Ads explained
FAQ
How much do OpenAI Ads cost?
It depends on how you buy. At launch in February 2026, reporting described roughly a $60 CPM (cost per 1,000 impressions) with six-figure entry minimums. Since then, CPC bidding arrived with observed bids around $3–$5 per click, and the entry minimum was dropped when OpenAI opened its self-serve Ads Manager to U.S. advertisers. For a B2B test, a defensible starting figure is around $10,000 to gather enough conversions to read the channel, but your true cost is measured in cost-per-qualified-pipeline, which depends on your funnel.
Is there still a minimum to advertise on ChatGPT?
The steep launch-era minimums came down. OpenAI opened a self-serve Ads Manager to U.S. advertisers and dropped the $50,000 minimum, which is what made the channel accessible to mid-market B2B teams rather than only large managed buyers. Always confirm the current entry requirements in the Ads Manager when you set up, since the program is still evolving.
Is CPA / conversion bidding available on OpenAI Ads?
Not as a mature, reliable self-serve bidding mode. OpenAI has publicly framed CPA bidding and third-party measurement as roadmap items it is promising and testing, and early cost-per-action formats are emerging, but you can't yet hand the platform your cost-per-acquisition target and trust it to optimize toward your CRM outcomes. For now, buy on CPM or CPC and attribute conversions yourself.
What's the difference between CPM and CPC pricing here?
CPM means you pay per thousand impressions regardless of clicks, good for reach and category awareness. CPC means you pay only when someone clicks through, better for demand capture and for modeling a cost-per-conversion. CPM was the launch model (~$60); CPC arrived later with observed bids around $3–$5 per click. Most B2B performance teams will start on CPC because it converts into a unit they can plan around.
Why is OpenAI Ads budgeting different for long B2B sales cycles?
Because the conversion you care about, a qualified demo or a closed opportunity, happens weeks or months after the click, usually inside your CRM. Mature platforms can optimize toward that with CPA bidding; OpenAI's measurement is still maturing, so the algorithm can't see your pipeline yet. That means you carry the conversion risk, you must instrument attribution yourself, and you should budget a learning period rather than promising a fixed cost-per-lead.
