On 24 August 2026 OpenAI began serving ads in ChatGPT across 31 European markets, its widest rollout of the format so far and roughly six months after the US pilot began. Two design decisions define it: ads appear only on the Free and Go tiers, and they launch without personalisation in order to comply with the GDPR.

What happened

Where. The 31 markets span the EEA and Switzerland, including Germany, France, Spain, Italy, the Netherlands, Belgium, Austria, Poland, Sweden, Norway, Denmark, Finland, Ireland, Portugal, Greece, Czechia, Hungary, Romania, and the Baltic states, plus Iceland, Liechtenstein and Switzerland.

Who sees them. Free-tier users and subscribers to the low-cost Go plan. Plus, Pro, Enterprise, Business and Education accounts remain ad-free.

How they target. They do not, initially. Ads launched in Europe without personalisation, which OpenAI attributes to GDPR compliance. Advertisers bought through major agency groups at launch, with self-serve access planned later.

Why unpersonalised matters more than it sounds

An unpersonalised ad in a conversational product is a genuinely different object from an unpersonalised banner. The conversation itself supplies context: what you asked about a minute ago is a far stronger commercial signal than a third-party cookie ever was. So the interesting question is not whether OpenAI uses a profile built from your history — it says it does not — but whether the current conversation is treated as targeting input, and under the GDPR whether that would need a lawful basis of its own.

This is the live regulatory question, and it lands in the same month the AI Act’s Article 50 transparency duties became enforceable . A user needs to be able to tell an answer from an advertisement. Where a commercial placement is generated or arranged by an AI system inside a conversational response, disclosure obligations and advertising law both apply, and neither is fully settled for this format.

Why it matters for builders

A monetised free tier changes what “free tier” means for your users. If your product embeds ChatGPT, or if your staff use free ChatGPT accounts for work, ads are now part of that surface in Europe. That is a procurement and acceptable-use question, not just an aesthetic one.

The tier split is the actual product signal. Enterprise, Business, Education, Plus and Pro stay ad-free. OpenAI is monetising the tiers it cannot monetise by subscription, and protecting the tiers that now out-earn consumer . If you are on a paid business tier, nothing changed for you — which is the point.

Watch for the commercial-content boundary in your own products. The wider lesson of 2026 has been that AI-mediated commercial content degrades in predictable ways when incentives and ranking interact — see Google’s Performance Max spam-lead problem and model collapse and AI slop . If you are building anything that mixes generated answers with paid placement, separate them structurally and label them, rather than relying on the model to keep them apart.

Personalisation will be the next announcement. Unpersonalised launches in Europe are typically a starting posture, not an end state. Whatever consent mechanism appears later is where the real privacy question gets decided — track it against AI privacy regulation .

Sources

  1. OpenAI, “ChatGPT Ads expands across Europe”: https://openai.com/index/chatgpt-ads-expands-across-europe/
  2. Euronews, “ChatGPT to start showing ads across Europe next week” (21 August 2026): https://www.euronews.com/next/2026/08/21/chatgpt-to-start-showing-ads-across-europe-next-week
  3. Digiday, “OpenAI’s ads business hits Europe at the six-month mark”: https://digiday.com/marketing/openais-ads-business-hits-europe-at-the-six-month-mark/
  4. MediaPost, “OpenAI Expands Advertising Pilot In 31 European Markets” (25 August 2026): https://www.mediapost.com/publications/article/417446/openai-expands-advertising-pilot-in-31-european-ma.html
  5. European Commission, “Safer and more transparent AI” (2 August 2026): https://commission.europa.eu/news-and-media/news/safer-and-more-transparent-ai-2026-08-02_en

Further reading