ChatGPT Is Now a Very Large Search Engine Under DSA Rules

ChatGPT is now regulated by what it does

I would not get too hung up on the word “search”: the key shift is that regulators are looking at a product’s actual function rather than the “chatbot” label. In an official decision, the European Commission classified ChatGPT as a very large online search engine under the DSA.

The threshold for this status is 45 million average monthly active users in the EU. At the time of classification, ChatGPT’s EU audience was estimated at roughly 159 million, putting it well above the limit. Once designated, the provider has four months to meet the additional requirements.

The list of obligations is substantial: assessing and mitigating systemic risks, undergoing independent audits, providing algorithmic transparency, and granting data access to vetted researchers. The assessment covers the spread of illegal content, risks to minors, fundamental rights, elections, and public safety.

For a search product, algorithmic transparency is especially sensitive. Questions about why the system chose a particular source, how it produced an answer, and which materials it effectively excluded now have a regulatory dimension. DSA violations can lead to fines of up to 6% of global annual revenue.

What changes for AI search and visibility

This is a genuine precedent for the entire class of AI assistants, not a narrow issue affecting one service. If Gemini, Claude, or another product performs a search function and crosses the 45 million EU-user threshold, a different category label will not shield it from similar classification.

For publishers and companies, the very unit of visibility is changing. In traditional search, you could track ranking position and clicks; in a generative response, what matters more is whether a source was selected, cited, or paraphrased. A user may receive an answer without clicking, while the wording of the original material is completely rewritten.

My main engineering question is now not only about ranking quality, but also about the verifiability of the selection chain. If AI search influences access to information, an opaque “the model decided” is no longer an adequate explanation. The most interesting part will begin where transparency requirements collide with closed models and constantly changing generation logic.

We previously covered how OpenAI API security measures can trigger alerts and require stronger compliance controls. That governance context becomes especially relevant as the EU applies DSA obligations to ChatGPT as a search engine.