What ChatGPT’s Self-Serve Ads Manager Means for Organic AI Visibility
On May 5, OpenAI opened a self-serve Ads Manager for ChatGPT. Any US business can now sign up at ads.openai.com, set a budget, upload an ad, and bid on a cost-per-click basis, with no minimum spend and no agency in the middle. Four months earlier you needed a six-figure commitment and a seat at the table with Dentsu, Omnicom, Publicis, or WPP to participate at all. The velvet rope came down fast.
I’ve been watching this rollout since January, partly because tracking AI citation behavior is what I do and partly because the question I kept getting from other practitioners was some version of “does this break everything we’ve been building organically?” A month into the self-serve era, I can give you my answer, the data behind it, and what I’d do differently between now and Q4 if I were running your AEO program.
How we got here, and how fast
The timeline here is important because the speed is the real story. OpenAI confirmed it would test ads on January 16. The pilot went live February 9 for Free and Go tier users in the US, with reported commitments around $200,000 and CPMs near $60, which priced out everyone except enterprise brands and their agencies. CNBC reported in March that the test was moving slower than the hype, with agencies frustrated that committed budgets weren’t being spent. Business Insider found advertisers getting weekly performance reports as CSV files, no dashboard, no real-time anything. Sensor Tower counted more than 100 brands advertising by spring, 44% of them retail.
Then May 5 happened. A self-serve Ads Manager in beta, CPC bidding added alongside CPM, a conversions API, a tracking pixel, and the minimum spend requirement gone. Ads now show to Free and Go users in the US, Canada, Australia, and New Zealand. Axios reported OpenAI is targeting $2.5 billion in ad revenue this year and $100 billion by 2030. A company doesn’t build a conversions API and a measurement pixel for a side experiment. Advertising is being wired in as a load-bearing part of the business, and every decision OpenAI makes about the chat interface from here on out happens with that revenue target in the room.
What the ads look like and how they’re targeted
The units themselves are restrained so far. Labeled sponsored cards sit below the model’s response, visually separated from the answer. A user comparing smartphones might see a Best Buy card; someone planning a weekend trip might get Expedia. Targeting runs on conversational context and chat history rather than keywords, with campaigns built in a familiar three-level hierarchy and pointed at “context hints” instead of search terms. Paid tiers stay ad-free, OpenAI says ads don’t influence the model’s answers, and conversations aren’t shared with advertisers.
The measurement side is young. AdExchanger covered the conversions API launch with the right amount of skepticism, since proving these ads work is the part OpenAI hasn’t done yet. Early click-through rates have reportedly landed below Google search benchmarks. If you’re used to the tooling in Google Ads, this will feel like 2003.
Whether paid placement touches the organic answer
The model generates its answer first, and the ad system appends a sponsored card after. Zen Media’s analysis of the rollout describes it the same way I’ve seen it behave: a brand can appear as a paid placement, as an organic mention, as both, or as neither, and the system doesn’t merge the two. Nothing in my own tracking suggests ad spend moves citations in either direction, and nobody publishing credible data has shown otherwise. I’d call that a fact for today and a hunch for next year.
The reason I hold the next-year part loosely is Google. Paid and organic started out separated there too, and the wall between them held for twenty years while ad load crept up the page, the visual distinction between ad and result thinned out, and organic listings slid below the fold on commercial queries. So the variable I’d watch in ChatGPT over the next year is how much surface area the organic answer keeps, because that’s what shrank at Google while everyone was watching the wall.
What keeps a brand in the answer
While the ad story was unfolding, AirOps published its 2026 State of AI Search report, and the persistence numbers in it should shape how you think about the organic side. Only 30% of brands stay visible from one AI answer to the next, and just 20% remain present across five consecutive runs of the same question. Pages that haven’t been updated quarterly are three times more likely to lose citations. Sequential headings and rich schema correlate with 2.8x higher citation rates, and brands earning both mentions and citations show a 40% higher likelihood of reappearing across answers.
Amsive’s research adds the stat I quote most often: brands are 6.5 times more likely to be cited through third-party sources than through their own domains. AI systems lean on what other people say about you far more than on what you say about yourself, which might be tough news for anyone whose content budget lives entirely on their own site.
What I’d do between now and Q4
Keep most of your budget off the paid side for now. If you run a test, run a small one with the conversions API wired up from day one, and treat the results as directional rather than decisive. The measurement tooling isn’t mature enough yet to justify big commitments, and early CPC competition is only going up from here. Treat your organic citation position as an asset that appreciated on May 5. Once a competitor can buy adjacency to the conversation, an earned mention inside the answer carries an implied endorsement that no sponsored card gets. The brands ChatGPT names on its own become the buyer’s shortlist, while the cards underneath get the same squint people learned to give sponsored results on Google.
I would definitely audit your coverage across the full cluster of questions a buyer asks, since both ad targeting and organic relevance follow conversations rather than keywords. A single optimized page covers one stop on a trajectory that might run six or eight questions deep.
Put a quarterly refresh cadence on the pages that earn your citations. The 3x citation-loss figure for stale pages is one of the cleaner cause-and-effect signals in the AirOps data, and a refresh calendar is cheap insurance. Shift real spend toward third-party presence with things like reviews, industry publications, communities, and the places your category gets discussed without you. The 6.5x stat says that’s where citation authority comes from, and it’s the one thing a competitor’s ad budget can’t buy quickly.
Money showing up in the channel was always going to happen, and the speed of this rollout tells you how much more is coming. What stays in your control is whether the model has a reason to name you when nobody’s paying it to. That’s the question I spent a whole book trying to answer, and the arrival of an ads auction next to the answer box makes it worth answering sooner rather than later.
Jarred Smith is the author of Explainable: Why AI Recommends Some Brands & Ignores Others, an Amazon bestseller on AEO, GEO, and SEO. He’s a marketing leader with nearly 20 years of experience across healthcare, public media, retail, and environmental services. Find him at jarredsmith.com.