ChatGPT Ads Are Live: What OpenAI's New Ad Platform Can (and Can't) Do Yet

Somewhere between the Google antitrust rulings and the latest Meta automation rollout, OpenAI quietly became an ad platform - and most media buyers still haven't opened Ads Manager to look. Three campaign objectives are live inside ChatGPT right now, conversion tracking runs through a real pixel and a Conversions API, and a batch of genuinely unglamorous advertiser tooling - budget pacing, geo exclusions, bulk campaign management - shipped in late July. This isn't a beta announcement anymore. It's an ad platform with a roadmap.

What You Can Actually Buy

ChatGPT Ads currently run on three campaign objectives, and the difference between them is the first thing to get straight before testing budget.

Tracking runs through a small SDK OpenAI calls OAIQ. It handles browser-side event collection and, the moment someone clicks an ad and lands on your site, sets a first-party cookie - __oppref - with a 30-day lifetime. There's also a server-side Conversions API for advertisers who don't want to rely on a browser pixel alone, which by 2026 is table stakes on every serious ad platform, not a nice-to-have.

Why This Isn't Just Another Ad Platform Launch

Two other moves this summer make it clear OpenAI isn't testing the waters - it's building infrastructure. It struck a deal with Yelp to pull in business listings, reviews, and photos for local recommendations inside ChatGPT, which is exactly the kind of first-party commercial data every ad platform eventually needs to make local and intent-based targeting work. And separately, Nielsen's $2.15 billion acquisition of DoubleVerify this August is the rest of the industry building the verification layer that any new ad-buying surface - ChatGPT included - will eventually be expected to support.

Put together: a real bidding model, real tracking infrastructure, real first-party commercial data, and an independent verification industry already pricing in the need to check its work. That's the checklist a platform completes on the way to being taken seriously, not the checklist of an experiment.

It's worth saying plainly that none of this guarantees ChatGPT Ads will become a top-five channel - plenty of well-funded ad formats never found an audience willing to click on them, and a conversational interface asking someone a direct question is a genuinely different environment for advertising to sit inside than a results page or a social feed. What the checklist does establish is that this is no longer speculative infrastructure. It's a functioning market with a bidding model, and functioning markets are worth a small, monitored position long before they're worth a real one.

One Compliance Wrinkle Before You Launch

If any of your spend touches European audiences, EU AI-disclosure rules now apply to ads delivered there - which matters more inside a conversational interface than it ever did in a search results page. A sponsored result sitting in a ranked list reads as an ad by convention. A sponsored answer sitting inside a conversation doesn't, unless it's labeled as one. Build the disclosure into your creative from the first campaign rather than retrofitting it after a compliance review flags it.

Should You Test It Yet

1. Test with a contained budget, not a reallocation. Treat it as incremental spend you can afford to lose, not a shift away from proven channels.
2. Wire up the Conversions API from day one. Beta reporting means you want your own source of truth running in parallel, not platform-reported numbers alone.
3. Start with the Conversions objective, not Views. Awareness spend is the hardest thing to justify on a platform with this little reporting maturity.
4. Reassess quarterly, not annually. A platform shipping this many features this fast will look different by year-end than it does today.

Nobody needs to move their budget into ChatGPT Ads this quarter. But the businesses treating this as a curiosity rather than a small, deliberate test are making the same call agencies made about Google Ads in 2001 and Facebook Ads in 2008 - mostly right in the short term, and expensive to have gotten wrong in the long one.