OpenAI flips the switch Monday across 31 markets. that's every EU country plus Norway, Iceland, Liechtenstein, Switzerland. Takes them from roughly nine ad markets to about forty in one go - more countries than every previous international launch combined.

the targeting is the story. No personalization at launch. OpenAI cites legitimate interest under GDPR for generic ads, says consent-based personalized targeting comes later after users get actively asked.

at launch you get: what the person is discussing, approximate location, device type, time of day, language. that's it.

One thing the user-side coverage misses: declining personalization doesn't remove ads. it just changes which ads appear. Getting rid of them entirely means upgrading.

Buying routes: OpenAI Ads Solutions team, agency partners, technology partners. self-serve Ads Manager is "later in summer 2026" with no date. Summer has five weeks left.

Once you're in: CPM, CPC, conversion-optimized CPC. country targeting. OpenAI Pixel, Conversions API, third-party integrations. Custom audiences appear in the general product description but assume they don't function in Europe at launch given the personalization constraint.

Everyone treats contextual-only as a downgrade. Don't agree.

A retargeting segment means someone visited a page days ago. a live conversation means someone is working through a problem in your category right now. That's a stronger intent signal, not a weaker one.

the cost is structural though. no sequencing. no frequency management across a funnel. No lookalike expansion. You're buying moments rather than people, which doesn't fit how most media plans, attribution models, or agency reports are built. Teams whose whole approach is funnel orchestration will find this unusable rather than just different.

Practical take: write the ten questions someone types immediately before they need what you sell. Those are your targeting units. Personas do nothing here.

Queue for self-serve now rather than when it opens. Test budget only.

What I'm unsure about: reporting suggests ChatGPT advertising hasn't been a runaway success in the US. six months in, nobody has published performance data convincing enough to justify moving meaningful budget. Whether ads inside a conversation feel helpful or intrusive at volume is a UX question six months of US operation hasn't publicly answered.

the specific targeting parameters for European advertisers come from press coverage, not OpenAI's advertiser docs, which I haven't been able to read directly. directionally right, not exhaustive.

Anyone running ChatGPT ads in the US or UK since launch? would like to see CPCs and conversion rates against comparable search spend. Haven't found anyone publishing real numbers.

a colleague on the organic side pointed out the language split - 31 markets, roughly two dozen languages, targeting keys off the live conversation. That conversation happens in Polish or Spanish, not English. Ask a question in Spanish and you get an entirely different set of cited sources, not translated versions of the English ones. Competition is far thinner. Most advertisers with real budget will start with English creative in the big five markets and leave the rest quiet.

Which makes the ten-questions exercise right but incomplete. Write them in the language the person would type them in. That isn't a translation job. People phrase the same problem differently, and the phrasing is the targeting.

Another person in the thread nailed it: moment-based targeting could outperform retargeting if you map out those trigger questions properly. Most teams just stare like you've got two heads.

for B2B clients we're pairing this with SEO-for-GPT tooling to see which prompts already surface them in ChatGPT, then using ads to amplify the high-intent ones and content to plug the gaps where competitors get named instead. The lack of personalization is the interesting part - a live conversation is way more telling than a cookie from three weeks ago.