OpenAI Added the UK to ChatGPT Ads but Skipped the EU on Purpose
OpenAI announced on May 7, 2026 that it is expanding its ChatGPT ads pilot to the UK, Brazil, Japan, South Korea and Mexico. Notably absent: every EU member state. The rollout adds five markets in weeks, none with a local third-party measurement partner signed. Self-serve buying remains US-only, which means holding companies own every new market by default.
That sequencing is the actual story. The press release frames it as "expanding thoughtfully" after "strong interest from businesses." Read against the absences and the gating, the expansion list says something more useful about how OpenAI is pacing its ad business and where the regulatory friction lives.
Five new markets, zero local measurement partners
The pilot launched in February 2026 and ran in the US, Canada, Australia and New Zealand for roughly three months. The new tranche brings the total to nine markets. Average monthly ad spend across the pilot has been about $109 million per Digiday's reporting, against a 2026 target of $2.5 billion.
Third-party measurement was promised at the US self-serve launch a few weeks ago. As of the May 7 expansion, OpenAI still doesn't have a partner signed or a timeline. What advertisers get instead is OpenAI's own Conversions API and a tracking pixel that reports OpenAI's counts back to OpenAI. We covered the deeper version of that gap when the US launch happened and it hasn't moved. Now it's true in five new currencies.
From what I've seen happen with every new ad inventory, the early advertisers in these markets will buy on platform-reported metrics for at least a quarter, possibly two, before any independent measurement is available. Whatever incrementality data they generate in that window is OpenAI's data, on OpenAI's terms. The CFO doesn't get a corroborating number from anyone else for a while.
Why the UK made the list and Germany didn't
The five expansion markets share something the EU doesn't: a tractable consent regime. The UK technically operates under UK GDPR after Brexit, but enforcement posture from the ICO has been noticeably lighter than the European Commission's, and the political appetite for AI-specific platform rules is lower. Brazil's LGPD allows opt-out interpretations US-style platforms have already adapted to. Japan and South Korea have lighter consent friction for ad measurement. Mexico's framework is closer to Brazil's than to anything in the EU.
Germany, France, Italy and Spain are absent because OpenAI hasn't finished building for them yet. Digiday reported in April that OpenAI's tracking pixel is being updated for EU compliance, with a new consent management system and a country field for jurisdictional data handling. That work isn't shipped. The Digital Services Act and the AI Act both apply on top.
So the rollout list isn't just "high ad-intensity markets," which is how the press release framed it via an analyst quote. It's the markets where the existing US infrastructure works with minimal rebuild. That detail matters because it sets a calendar. EU advertisers should probably not expect access this calendar year, and possibly not in the first half of next year either, depending on how the AI Act enforcement schedule plays out. UK advertisers got access first because the UK has, in practice, become OpenAI's rest-of-Europe testing ground for anything regulated.
The hiring map backs this up. OpenAI is staffing ads executives in London and Tokyo, per Digiday. So the UK and Japan picks aren't just regulatory convenience, they're also where local teams are getting built first. That's a strong signal of where the next wave of self-serve and measurement work will land geographically before it lands in the EU.
Self-serve gating hands these markets to four holding companies
Here is the quietest part of the announcement. Self-serve ads manager access stayed US-only. UK, Brazil, Japan, Korea and Mexico advertisers can only buy through OpenAI's named agency partners: Dentsu, Omnicom, Publicis and WPP. The CPC bidding mode that opened up in the US a couple of days ago is also a US-only feature for now.
Functionally, this means a London brand that wants to test ChatGPT ads in May 2026 has to go through one of four holding companies. There is no direct buying surface. There is no way to bring an independent agency in and run small tests. Minimums are whatever the holding company sets, markup is whatever the holding company sets, and data flows through whatever measurement stack the holding company uses.
If you are an indie agency or an in-house team in any of the five markets, the practical implication is that ChatGPT ads aren't really an option for you yet. They're an option for clients big enough to be sitting inside Dentsu, Omnicom, Publicis or WPP's account roster. That's a real distortion. It's also what every gated ad rollout has looked like in the first six months. The question is how long it lasts before self-serve opens.
The categories where this gap matters most
OpenAI's stated focus categories for the expansion are shopping, retail and travel. Those are exactly the three verticals where post-click attribution is most valuable, because the conversion is a real purchase and the deltas between platforms are how budget gets reallocated quarter to quarter.
Buying performance inventory without independent measurement is a defensible call when you are testing a new channel with discretionary budget. It's a less defensible call when the channel is ranked against Meta Advantage+ or Google AI Max in your quarterly review. There are buyers in every market who have been here before, with TikTok in 2020 and with Reels Performance Ads in 2023. The pattern has been consistent: incrementality looks great when measured by the platform, ambiguous when measured externally, and the truth shows up in a third-party audit a year later.
So a UK or Japan retail advertiser piloting ChatGPT ads this quarter is in a familiar early-stage spot. The ROAS will read well in OpenAI's interface. CPC pricing made the unit economics easier to model than under the original CPM. Independent corroboration is still the missing piece, and it will stay missing for a while.
Three things to lock in before money moves
If you are working an account in the UK, Brazil, Japan, South Korea or Mexico, three things are worth doing before any spend actually moves.
First, get the conversion lift question on paper now. Decide what proof of incrementality you would accept. A geo holdout. A time-based on-and-off test. A matched-market control. Whichever you can run cleanly in your shop. Without that pre-commitment, the platform-reported numbers will be the only numbers in the room when budget decisions happen six months from now.
Second, ask your holding company partner what their measurement layer looks like and whether they have a path to independent verification that doesn't terminate at OpenAI's pixel. If the answer is "we use OpenAI's reporting," you have your answer about how this story looks in two quarters.
Third, watch the EU launch date as a signal. Whenever Germany and France come online, expect that OpenAI will have shipped the pieces missing today: consent management, jurisdictional data handling, and probably a third-party measurement partner because the EU's regulatory pressure forces it. That release will be more advertiser-friendly than what UK and Brazil get to start with. Plan accordingly. The right time to put real budget into these markets is probably after the EU launch, not before.
The expansion doesn't change the pilot's central feature, which is that OpenAI is building the inventory and selling the inventory and counting the inventory all in the same room. It just adds five new countries where that's the deal you take if you want to be early.
Notice Me Senpai Editorial
