ChatGPT Ads are already a European advertising channel.
The harder question is what happens when an ad system moves beyond the context of the conversation happening now and starts using what a person has told an AI assistant over time.
That question moved closer to the centre of European advertising policy this month.
On 2 October, North Rhine-Westphalia’s data-protection authority warned about planned broader ChatGPT ad profiling in Europe. The authority described a model in which previous and current chats, device information and usage behaviour could contribute to advertising profiles, subject to opt-in consent.
That was advice from a German regulator, not an enforcement decision against OpenAI.
It also does not mean profile-based ChatGPT advertising is already active throughout Europe.
OpenAI currently says personalised ads are not initially available in the European Economic Area or Switzerland. Ads can still be selected using the context and intent of the current chat.
For marketers, that distinction matters more than simply asking whether ChatGPT Ads are live.
Ads availability and ad personalisation are different feature states
OpenAI announced ChatGPT Ads across 31 European markets in August and made self-service access through Ads Manager available on 31 August.
NEMO covered the buying, measurement and commerce implications in our guide to ChatGPT Ads in Europe.
But access to an advertising product does not establish that every targeting or personalisation capability is available in every market.
OpenAI’s current advertising documentation says its system can use the context and intent of the current conversation when selecting an ad.
Where ad personalisation is enabled, additional signals from a user’s wider ChatGPT experience can also contribute. OpenAI says those signals may include past chats and memory.
The European caveat is important: personalised ads are not initially available in the EEA or Switzerland.
OpenAI also says advertisers do not receive people’s ChatGPT conversations, chat history, memories or personal details. Those signals, where eligible for use, stay within ChatGPT’s advertising system.
So the accurate European framing is not that advertisers are receiving chat histories.
It is that chat-derived signals can become inputs into ad selection when personalisation is enabled, and European regulators are already scrutinising how that transition should work.
Germany’s regulator is warning before the wider feature arrives
North Rhine-Westphalia’s data-protection authority said on 2 October that more extensive profile-based ChatGPT advertising was expected in Europe.
Its description separates two models.
One is contextual advertising based primarily on the conversation happening now and limited contextual information.
The second is broader profile-based advertising drawing on a longer history of interaction. The authority says that model should depend on opt-in consent and advised users against agreeing to it.
The statement does not establish an infringement.
No fine was announced. No ban was announced. NEMO has not identified a formal enforcement proceeding against ChatGPT Ads arising from the statement.
But it is still operationally important for advertisers because European media plans can easily assume that a feature available elsewhere will produce the same addressable audience inside the EEA.
That assumption is unsafe here.
Cologne has already ruled on another chatbot-advertising setup
A separate German case provides a more concrete warning.
On 17 September, the Cologne Regional Court ruled in a case brought by Germany’s consumer federation against Snap Group Limited over Snapchat’s My AI chatbot and related advertising settings.
The court ordered Snap to stop the advertising-data processing challenged in the case where there was no valid legal basis.
It also rejected pre-ticked advertising-preference boxes as a valid consent mechanism.
A further part of the ruling concerned accounts belonging to minors and the pre-selection of advertising interests including alcohol and gambling.
The case matters because chatbot conversations can contain information that is very different from an ordinary product-page visit.
People may discuss health, religion, political views, finances, relationships or other highly personal subjects with an AI assistant. The Cologne court considered the possibility that chatbot inputs could include sensitive information even where the platform did not intend to collect that category of data.
That does not mean chatbot data can never be used for advertising.
It means Snap’s specific implementation, consent mechanism and legal basis failed in the setup considered by the court.
The ruling is also a first-instance German judgment, not an EU-wide precedent. That is the same guardrail NEMO applies in our coverage of the Munich AI Overview case.
Four signals advertisers should keep separate
| Signal | What it establishes | What it does not establish |
|---|---|---|
| European ChatGPT Ads rollout | Advertisers can buy ChatGPT inventory in supported European markets | That all personalisation features are available everywhere |
| OpenAI EEA documentation | Personalised ads are not initially available in the EEA or Switzerland | That personalisation will never launch there |
| NRW regulator statement | A German DPA is publicly warning about broader chat-based profiling | That OpenAI has been fined or found unlawful |
| Cologne My AI judgment | Snap’s challenged advertising setup lacked a valid legal basis | That all AI-chat advertising is unlawful |
The common mistake would be to collapse all four into one claim about “AI advertising regulation.”
They are different layers.
One concerns product availability.
One concerns platform design.
One is regulatory guidance.
One is a court judgment involving a different company.
Run this European chat-advertising check
Before scaling spend on conversational advertising, European teams should document what they are actually buying.
1. Record the market and feature state.
Do not write “ChatGPT Ads available” and “ChatGPT personalised ads available” as if they mean the same thing.
2. Separate current-chat context from persistent profiling.
Ask whether ad selection is based on the immediate thread or a broader history of user signals.
3. Audit the advertiser-side data flow.
If your team uses pixels, conversion APIs or custom audiences, document the information your business supplies and the legal basis for sending it.
4. Test audience assumptions.
Do not forecast an EEA campaign using an addressable audience that depends on personalisation currently unavailable in the region.
5. Review sensitive-category adjacency.
AI conversations can contain unusually personal information. Treat health, finance, religion and similar topics as a brand-safety and governance question, not just a targeting opportunity.
6. Review age controls.
The Cologne case shows why alcohol, gambling and other restricted categories need explicit attention where minors may use the underlying product.
7. Keep contextual and personalised performance separate.
If broader personalisation later becomes available, measure it as a change in the media environment rather than silently mixing results with contextual inventory.
The EU’s wider AI rules create a separate compliance layer. NEMO’s EU AI Act Article 50 marketing checklist covers transparency requirements that should not be confused with the GDPR and advertising-consent questions at issue here.
What NEMO will watch next
The next meaningful development is not another broad statement that AI advertising is changing.
It is evidence that the European feature state has changed.
NEMO will watch for an actual launch of personalised ChatGPT Ads in the EEA, the consent interface used for that launch, any position from Ireland’s Data Protection Commission, and any appeal or later development in the Snap case.
Until then, marketers should use a simple rule:
ChatGPT Ads being available in Europe does not mean chat-history-based ad personalisation is available in Europe.
That distinction belongs in media plans, measurement documentation and regulatory reviews now.