ChatGPT ads are live. Here's how to test without getting taken.
Ads inside ChatGPT are running now - live in the US, Canada, Australia, and New Zealand, with more markets opening. No minimum spend, barely any benchmarks, and a window that won't stay quiet for long.
Every few years a new auction opens and, for a while, attention costs less than it's worth. Search in the 2000s. Facebook in the early 2010s. TikTok more recently. Now it's ads inside ChatGPT - and unlike most "next big things", this one is already taking budgets in four countries.
The window is real but it has a shape: no minimum spend, thin competition, and almost no reliable benchmarks. That last part is why early advertisers either learn cheaply or overpay quietly. Nobody selling a mature ChatGPT ads playbook has one - the platform is months old. What we have is a repeatable way to enter new channels, tested on every previous one.
How we run it
- Live markets: test now: US, Canada, Australia, New Zealand, with the UK and more following. Structured pilot, capped budget, kill criteria agreed before any budget is spent.
- Waiting markets: get ready: accounts, creative, tracking, and the organic groundwork prepared so launch week isn't spent on admin.
- Organic and paid together: ads put you next to the answer; the AI visibility work puts you in it. Running one without the other wastes half the effort.
- Measurement that doesn't flatter: new platforms grade their own homework generously. We reconcile against analytics and pipeline before believing anything.
- A straight recommendation: for some businesses the right call is still "watch, let others pay the tuition." If that's you, we'll say so. It's cheaper to be told early.
The advertisers who did best on every previous platform showed up early and kept their discipline. Showing up early without it is just a faster way to spend the budget.
Questions we get asked
Are ChatGPT ads actually available?
Yes - this stopped being a preview in 2026. A US pilot opened in February, self-serve access followed for all US advertisers in May, and Australia, New Zealand, and Canada went live in April, with the UK following and Mexico, Brazil, Japan, and South Korea announced. No minimum spend. If you're in a live market, the only queue is your own readiness; if you're not, the preparation work means you move the week your market opens.
What do the ads look like?
Sponsored cards under ChatGPT's responses, shown to logged-in users on the free and lower-priced tiers - paying subscribers on premium plans don't see them. Formats and policies are still moving, so treat any specifics as dated the month they're written, including ours.
Why advertise somewhere with no benchmarks?
Because no benchmarks cuts both ways: nobody knows what a fair price is, and early auctions on every platform in history have under-priced attention - we've written about what happened when those windows closed. The discipline is a capped budget, pre-agreed kill criteria, and measurement you can check. Same rules we apply to any emerging channel.
Should we do this before or after the organic AI work?
Alongside, ideally - they compound. An ad might win the click, but the assistant's underlying recommendation of your brand comes from the organic visibility work: entities, citations, structured content. Paying to appear next to an answer that recommends your competitor is a bad look.
How do you measure ads in a channel this new?
Conservatively. Platform-reported numbers get treated as claims, not facts - we verify against landing traffic, branded search lift, and pipeline, the same way we keep every other platform honest. If the channel can't show incremental value inside the test window, we say so and stop.
Related reading
Want this handled properly?
Tell us where you're at. We'll tell you what we'd do - and what it costs.