OpenAI Expands ChatGPT Ads to 31 Countries in Europe

OpenAI just gave European ChatGPT users something new to scroll past: ads. As of August 24, 2026, ChatGPT Ads is live across 31 countries in the European Economic Area plus Switzerland, the platform's biggest expansion since it launched in the US back in February. If you're the engineer who ends up owning the ad pipeline once marketing says "we're doing this," this is the week to check your integration.

What launched in ChatGPT ads Europe

Ads show up only for people on the Free and Go plans. Plus, Pro and Enterprise subscribers still get an ad-free ChatGPT, and OpenAI says the ads are "clearly labeled and separate from ChatGPT's answers" and don't influence what the model tells you. Advertisers can buy in through OpenAI's Ads Solutions team, an agency, or a technology partner, and self-serve access through Ads Manager has been rolling out alongside it, now live in more than 50 countries including most of the EEA.

Two ad types, two legal bases

The part that should actually change what you build is the consent model. Under GDPR, OpenAI splits ads into two tiers:

• Contextual ads use only the live conversation topic, approximate city-level location, device, time of day and language. These run without an opt-in prompt.
• Personalized ads draw on ad interaction history, inferred interests and advertiser conversion data. These require explicit, revocable opt-in consent, stored per account and changeable anytime in settings.

That's a real departure from the US rollout, which runs on an opt-out model. Opting out of personalization in Europe doesn't turn ads off, it just drops the user back to contextual-only.

Why this matters for the pipeline, not just the campaign

If your team is wiring conversion tracking into ChatGPT Ads, the consent tier a user falls into decides what data you're even allowed to send. OpenAI's server-to-server Conversions API expects hashed identifiers, not raw ones:

curl -X POST "https://bzr.openai.com/v1/events?pid=<PIXEL-ID>" \
  -H "Authorization: Bearer <API-KEY>" \
  -H "Content-Type: application/json" \
  -d '{
    "validate_only": false,
    "events": [
      {
        "id": "order_12345",
        "type": "order_created",
        "timestamp_ms": 1773892800000,
        "action_source": "web",
        "user": {
          "emails_sha256": ["<sha256-of-lowercased-email>"],
          "countries": ["DE"],
          "ip_address": "203.0.113.1",
          "user_agent": "Mozilla/5.0"
        },
        "data": {
          "type": "contents",
          "amount": 2599,
          "currency": "EUR"
        }
      }
    ]
  }'

Every identifier in the user object has to be SHA-256 hashed before it leaves your server, geographic fields are the only exception, and a batch of up to 1,000 events fails entirely if one event in it is malformed. Send personalization-derived fields for a user who never opted in, and you've turned a tracking bug into a compliance problem for whoever owns the data processing agreement with OpenAI.

What to actually do Monday morning

Check three things in your Conversions API integration: whether it's hashing identifiers correctly, whether personalized-event fields are gated behind a real consent check instead of a flag that defaults to true, and whether your EU test traffic is showing up in Ads Manager at all yet. OpenAI's US pilot saw CPMs drop from around $60 to the $25 to $45 range as the auction opened up to more advertisers. Expect European pricing to move the same way once self-serve volume builds, so don't lock in a bidding strategy off the first week of data.

The bigger point: this is the first time a chat interface, rather than a search page or a social feed, is running consent-gated advertising at this scale. If you're the one building the plumbing between a conversational product and an ad network, the GDPR split between contextual and personalized signals is a pattern you'll be implementing again elsewhere.