OpenAI’s self-serve ads turn ChatGPT into a media channel
OpenAI just opened self-serve ChatGPT Ads with CPC bidding and conversion tracking. The GEO industry's measurement story just got harder to defend.
OpenAI quietly opened self-serve access to ChatGPT Ads yesterday. CPC bidding, Conversions API support, pixel-based tracking, campaign pacing, the lot. US advertisers can register now; the rest of us are next.
This is the bit most people will skim past. They shouldn't. The pilot phase — where you needed an agency relationship with Dentsu or Publicis to even get a meeting — is functionally over. ChatGPT just became a media buying platform you can log into on a Tuesday afternoon and spend money on.
Similarweb's data, also published this week, says ChatGPT ad CTR averages 0.68%, with the best brands hitting 1.57% and a peak of 5.4%. Those aren't display-network numbers. Those are search-adjacent numbers. Which is exactly what they should be, because that's exactly what this is.
The thing nobody is saying out loud
For the last eighteen months the GEO industry has been selling clients on a story where ChatGPT is a discovery layer you optimise *for* — get cited, get mentioned, get into the answer. It's a polite story. It's also incomplete.
The complete story is that ChatGPT is now a discovery layer you can also *buy your way into*, with CPC bidding and conversion tracking, on a self-serve dashboard that looks suspiciously like Google Ads circa 2008.
Google's organic-paid duopoly took fifteen years to settle into the shape we know. ChatGPT is going to do it in about three.
What the CPC pivot actually signals
OpenAI launched its ad pilot on CPM. Switching to CPC isn't a feature update — it's a positioning move. CPM is for brand advertisers buying eyeballs. CPC is for performance advertisers buying intent.
The honest GEO playbook just acquired a paid leg, and most strategy decks haven't been rewritten yet.
OpenAI knows the second pile of money is bigger than the first. They've watched Google print it for two decades. The CPC pivot tells you who they're actually building this for: in-house performance teams, lead-gen businesses, e-commerce, SaaS — anyone who already runs Google Ads and is now sitting on a board deck saying "we need an AI search strategy."
Those advertisers are not going to fund a six-month GEO retainer to maybe get cited. They're going to put a card on file and bid on relevant queries, because that's what they already know how to do.
The honest GEO playbook just acquired a paid leg, and most strategy decks haven't been rewritten yet.
The measurement asymmetry is about to flip
For the last year I've been arguing that the measurement problem is the actual problem with AI search. Nobody can prove what's working. Citation tracking is immature. Server logs are inconsistent. The whole industry is operating partially blind.

Paid changes this. Conversions API, pixels, downstream attribution — these are mature systems. Imperfect, sure, but battle-tested. Within a few quarters, performance marketers running ChatGPT Ads will have cleaner numbers on AI-search ROI than anyone running an organic GEO programme.
That asymmetry is going to be uncomfortable. The CMO sitting in a quarterly review will see two columns: one with CPCs, conversions, and a CAC figure from ChatGPT Ads, and one with "estimated citation share" and "brand mentions in LLM answers" from the GEO retainer. Guess which line item gets defended at the next budget meeting.
The organic side has eighteen months, maybe less, to develop measurement that survives that comparison. Brand-lift studies, log-file analysis, share-of-model voice — something concrete. The "we can't really measure it but trust us it's working" pitch has a shelf life and the timer just started.
Why this lands harder for small businesses than the industry assumes
The reflex take is that self-serve ChatGPT Ads is a big-brand story. Wrong, I think. Self-serve buying is specifically what unlocks small and mid-sized advertisers, because they were the ones priced out of the managed-partner model. A plumber in Birmingham was never going to retain Omnicom to buy ChatGPT inventory. A plumber in Birmingham can absolutely log into a dashboard and bid on "emergency boiler repair near me" if the queries route through ChatGPT.
This is the same dynamic that made Google Ads dominant. Self-serve democratised access, and the long tail of small advertisers ended up generating the bulk of the revenue. OpenAI knows this. The Ads Manager isn't being built for WPP. WPP already has a phone number.
What it means for UK service businesses specifically: the window where you could ignore ChatGPT as a paid channel because "it's not really available yet" closes in roughly the time it takes the beta to expand internationally. Probably six to nine months. After that, your competitors will be bidding on your category and you'll be deciding whether to follow.
The MacIsaac lawsuit in the same news cycle is not unrelated
Worth noting that the same week OpenAI is opening self-serve advertising, Google is being sued over a $1.5m AI Overview defamation claim from a Canadian musician falsely identified as a sex offender. The two stories are connected, even if nobody's drawing the line.
The line is this: AI answer surfaces are now commercial real estate. They're being monetised, ranked, and increasingly held legally accountable. The era where "the AI just made a mistake" was a defensible position is ending. Once you sell ad inventory adjacent to AI-generated answers, "we're not really a publisher" gets harder to argue. OpenAI is walking into the same regulatory and legal exposure Google has spent twenty-five years navigating, and they're doing it at speed.
For advertisers, the practical implication is mild but real: brand-safety questions in AI answer environments are going to come up. The industry doesn't have answers yet. Worth asking now rather than after the first big incident.
What to actually do about this
If you run a UK business that already runs Google Ads: register for ChatGPT Ads access now, even if you can't buy yet. Get in the queue. When access opens, allocate a small test budget — 5-10% of your search spend — and treat it as a learning channel for two quarters. Don't expect Google-level scale. Do expect cleaner intent data than you're getting from most social platforms.
If you run an SEO or GEO retainer: your clients are going to ask about this within the next month. Have a position before they do. The wrong position is "ignore it, organic is what matters." The right position is "here's how paid and organic work together in AI search, and here's what we do on each side."
If you sell GEO services as a standalone discipline: the ground just shifted. The argument that AI search is a separate channel requiring separate expertise was always shaky. It's about to face a much harder test, because performance marketers are about to start treating it as a paid search channel with a different inventory partner. Your positioning needs to account for that.
The measurement gap, the brand-versus-paid balance, the question of who owns AI search inside the organisation — all of these were theoretical six months ago. They're now operational. The businesses that figure out the answers first will have a quiet but real advantage over the ones still running playbooks written for a world where ChatGPT didn't sell ads.
That world ended yesterday.
Ready to improve your visibility in AI search?
If you're an SME in Surrey or London and you want more qualified leads from search — including the growing AI answer layer — let's talk.
Book a discovery call