ChatGPT ads in 2027: how to budget, test and contract for them
ChatGPT ads now run on a real auction, but measurement is young. How to budget, test, measure and contract for them, and who should own the channel.
ChatGPT ads should be budgeted and run like a new search channel: a fixed test budget, a written hypothesis, holdout-based measurement and a clear owner inside your existing search or agency scope. The auction is real now, but the measurement is young. Treat early results as directional, contract for transparency up front, and don't let the spend turn into a line nobody can explain to the CFO.
Is ChatGPT advertising a real channel yet?
Yes. A year ago it was a premium beta. Today it's an auction.
The pilot launched in the US on February 9, 2026, in the Free and Go tiers. Early buys were sold on impressions at a reported $60 CPM with commitments around $200,000. Within months, self-serve Ads Manager arrived, the minimum fell to $50,000, cost-per-click bidding was added, and reported CPMs dropped as low as $25.
The mechanics now look familiar to any search team. OpenAI is testing multi-advertiser ad units, where several relevant ads share one placement, priced through a second-price auction. Impression campaigns support custom CPM max bids. Ads Manager added daily budgets, one-click cloning from CPM to CPC, bulk editing, and average daily budgets with weekly pacing. Targeting has expanded beyond the US, Canada, Australia and New Zealand to the UK, Japan, South Korea, Brazil and Mexico.
Scale followed. By September, Channel Insider reported ChatGPT ads had passed a $1 billion annualized run rate in under 200 days, with tens of thousands of advertisers, availability in more than 40 countries and more than 50 technology and measurement partners. Axios reported OpenAI told investors it expects $2.5 billion in ad revenue this year. Most campaigns, per Axios, run on CPC or outcome-optimized bidding.
That's a channel. It's also a channel your competitors are already bidding in.
How much should we budget for ChatGPT ads in 2027?
Budget it as a test, not an allocation. The question for 2027 isn't "what share of search goes to ChatGPT." It's "what do we need to spend to learn whether this works for us."
A sensible structure:
- Fund it from search, not from nowhere. People now ask AI assistants what they used to type into Google. Omnicom research put generative AI at nearly a quarter of searches in 2025. If ChatGPT ads capture the same intent, the money should come out of the same pool and be judged against the same cost per outcome.
- Size it to reach a readable result. A test that's too small to produce a conclusion is just a donation. Work back from your conversion rate and the volume you need to compare against a control.
- Set a stop date and a stop rule. Decide in advance what cost per acquisition kills the test and what result earns more budget.
- Keep it a named line. ChatGPT ads should have their own budget code, not hide inside "paid search other" or a platform-fees bucket.
How do we measure something this new?
Carefully, and with your own controls. Measurement is the weakest part of the channel today, and that's normal for a platform this young.
The pieces are arriving. In June, LiveRamp became the first independent ad tech company connected to OpenAI's conversion API, starting with offline transaction data in the US for mutual clients. Useful. But Digiday also noted that OpenAI decides how matching and attribution work, and one consultant warned that partners with commercial ties to the platform risk "grading their own homework." LiveRamp is also due to become part of Publicis, which is worth watching if your agency sits in a rival holding company.
What we'd put in place:
- Holdouts or geo tests you design, not only platform-reported conversions.
- Incrementality against search. If ChatGPT ads claim a sale your brand search would have won anyway, you've paid twice.
- First-party conversion data sent through a conversion API, so you're not relying on clicks alone.
- A finance sign-off on the method before the test starts, not after the results arrive.
Who should own ChatGPT ads: agency, search team or AEO?
The search team owns the bidding. The agency runs it inside existing scope. AEO sits next to it, and someone has to connect the two.
ChatGPT ads and answer engine optimization (AEO, being the brand an AI recommends in its organic answers) live on the same screen. A paid placement next to an answer that leaves you out tells you something. An answer that already recommends you often means the ad buys little. Splitting the two across teams that never compare notes is how money gets wasted. We made the same point in our piece on what changed in media agency reviews for 2027: AEO usually has no clear owner, so put it in scope.
In practice that means one person accountable for AI visibility, paid and organic, with a monthly view of both. It can sit in-house or with the agency. It can't sit nowhere.
What should the agency contract say?
Most agency contracts were written before this channel existed. Before your agency spends a dollar here, get these on paper:
- Scope and fee. Is ChatGPT ads covered by the current search fee, or billed separately? If separately, at what rate and for how long? A new channel is the easiest place for a quiet fee increase.
- No markup on media. Spend should pass through at cost, billed in the advertiser's name or with platform invoices shown.
- Account ownership. The Ads Manager account, conversion API connection and data belong to you.
- Reporting. Spend, bids, placements and conversions reported line by line, with the test design and results shared in writing.
- Partner disclosure. If the agency's holding company has a commercial relationship with a measurement or data partner, you hear about it.
For the wider commercial picture across agencies, AI tools and platforms, see how we approach marketing spend and AI spend.
What's the right next step?
Pull three things before the 2027 plan locks: your current search agency scope of work, any ChatGPT ads spend already running (check the card statements, not just the media plan), and the measurement method your finance team will accept. Then write a one-page test plan with budget, owner, stop rule and contract terms. If you'd like an independent read on the agency terms before you sign, our Second Opinion reviews a single contract in five business days.
Sources: eMarketer, ChatGPT's $200K minimums for ad beta signal OpenAI's push to monetize scale, PPC Land, ChatGPT ads finally leave the US: UK, Japan, Korea, Brazil and Mexico next, Search Engine Land, OpenAI to expand ChatGPT ads to new markets, test multi-advertiser placements, Search Engine Roundtable, ChatGPT multi-advertiser placements and Ads Manager updates, Channel Insider, ChatGPT Ads reach $1 billion as OpenAI expands self-service access, Digiday, OpenAI's ChatGPT ads get its first conversion API partner in LiveRamp, Digiday, AI's impact on search forces agencies to rewrite the marketing script
Are ChatGPT ads worth testing in 2027?
Yes, for most brands with meaningful search spend. The channel now runs on an auction with CPC and CPM bidding, so a controlled test with a fixed budget and stop rule is the right way to find out if it works for you.
How are ChatGPT ads priced?
Early ads were sold on impressions at a reported $60 CPM. Advertisers can now bid on CPM or CPC through Ads Manager, and OpenAI is testing multi-advertiser placements priced through a second-price auction.
How do you measure ChatGPT ad performance?
Use your own holdout or geo tests alongside platform reporting, and send first-party conversions through a conversion API. Agree the method with finance before the test starts.
Should ChatGPT ads be part of our search agency's scope?
Usually yes, with the fee, account ownership and reporting written down. Pair it with whoever owns AEO so paid and organic visibility in AI answers are judged together.
That is usually the first sign there is money in it. Fifteen minutes and a recent media invoice is enough for us to tell you where we would look first.