OpenAI’s Ads Manager update is the story I/O obscured

OpenAI quietly shipped state, DMA, and zip code targeting in its Ads Manager Beta this week, alongside daily budgets and aggregate reporting totals. The coverage treated it as a routine product update. It isn't.

What OpenAI shipped is the location-targeting feature set that took Google AdWords years to build out. They shipped it inside the platform that's pulling search demand away from Google. And they shipped it the same week the industry was busy arguing about whether Google's AI Overviews count as "the end of search."

The conversation is on the wrong story.

What actually changed

The headline features are mundane in isolation. Daily budgets instead of lifetime-only. State, DMA, and zip code targeting in the US. Aggregate totals in the reporting tables. Dynamic CTAs — "Shop Now", "Book Now", "Sign Up", "Learn More" — being tested inside ChatGPT responses.

OpenAI isn't building an ad product. They're building the second Google.

Read individually, each of those is a checkbox. Read together, they're a platform crossing a threshold. Daily budgets and granular geo targeting are the table stakes for a paid media channel to be taken seriously by performance teams. Without them, ChatGPT Ads was a brand experiment. With them, it's a media buy you can actually plan against a forecast.

The CTA test matters more than the geo update, though it's getting less attention. Dynamic calls-to-action selected based on creative and destination are the mechanism by which an AI assistant becomes a conversion surface. Not a discovery surface. Not a brand surface. A place where a user clicks "Book Now" and something measurable happens on the other end.

OpenAI isn't building an ad product. They're building the second Google.

Why this is bigger than I/O

Last week's Google I/O announcements got the headlines. The new search box that accepts images, files, videos, and tabs. AI Mode at a billion monthly users. Information agents that monitor the web on behalf of users. TechCrunch declared search dead. Time warned of disruption. LinkedIn did its usual thing.

But Google's I/O announcements were about defending an existing position. The new search box is a UI evolution of a product that's already the default. The agents are extensions of a search habit that already exists. Even Liz Reid's messaging the day after — "AI Mode is not the default experience in Search" — was the language of an incumbent reassuring the market that nothing has fundamentally changed at the centre.

OpenAI's Ads Manager update is the opposite kind of announcement. It's the language of a challenger building out the missing infrastructure to compete for the same dollars. And the dollars are what determine which discovery surface gets prioritised by businesses over the next three years.

If you're a UK service business, you don't actually care which AI assistant your customers use. You care which one you can buy targeted distribution on. Until last week, the honest answer was Google. From this week, the answer is starting to be "Google, and also OpenAI, with caveats."

The measurement problem just got harder

I've written before that the AI search measurement problem is the actual problem — that until log analysis, server-side tracking, and citation monitoring mature, everyone is partially blind on AI performance. The Ads Manager rollout makes that problem worse in a specific way.

When ChatGPT was purely an organic discovery surface, the measurement question was "are we getting cited?" Hard to answer, but bounded. Now there's a paid layer being added on top of the same surface. So the measurement question becomes "are we getting cited, are we getting clicked from a paid placement, and how do those two interact?"

This is exactly the problem Google solved over fifteen years by building Search Console, Google Ads reporting, and GA in a tightly coupled stack. OpenAI's Ads Manager has aggregate totals in a table view. That's a start. It's not a measurement stack.

The honest position right now: if you're spending on ChatGPT Ads, you can see what you spent. You can see clicks. You can see impressions. You can't yet see how that paid activity interacts with the organic citation behaviour of the same model, because the organic citation behaviour of the same model isn't measurable in the first place.

Anyone selling you a unified AI search dashboard right now is either selling you log analysis with a markup, or selling you a number they invented.

What this means for UK businesses

Three things to actually do, all of them undramatic.

First: if you run paid acquisition and you haven't yet looked at ChatGPT Ads, the geo-targeting update is the moment where it becomes worth a small test budget. Not a media plan rewrite. A test budget. State-level US targeting is live now; international rollout will follow the same pattern these platforms always follow, and you want to have ad ops familiar with the interface before it matters to your market.

Second: separate your thinking about ChatGPT-as-discovery from ChatGPT-as-paid-channel. They're the same surface for the user, but they're different problems for you. Discovery is about whether your content and brand earn citations. Paid is about whether you can buy targeted impressions. Conflating them — which most of the GEO consulting industry currently does — leads to strategy decks that try to solve both with the same tactics.

Third: stop letting Google I/O announcements set the tempo of your AI search planning. Google announcing things is not the same as the AI search landscape shifting. The actual shift this month was OpenAI building the second functional ads platform, in public, while everyone watched Mountain View.

The platform that gets boring wins

Most product updates in advertising are boring. Daily budgets. Aggregate totals in a reporting view. DMA targeting. These are the unsexy features that mature platforms have and immature platforms don't. The boring features are what make a platform usable by media buyers at scale.

OpenAI getting boring is the most significant thing it could be doing right now. Boring means buyers can plan. Boring means agencies can build practices around it. Boring means CFOs can sign off on the budget. Boring is what turns a novelty into infrastructure.

Google spent two decades being the boring platform you couldn't ignore. The bet OpenAI is making — and the I/O panic obscures it — is that they can become the second boring platform you can't ignore, and they can do it in three years instead of twenty.

The geo-targeting update isn't the news. The news is that updates like this are now arriving on a monthly cadence from a platform that didn't have an ads product eighteen months ago. That's the loop. And it's accelerating.