The EU has decided ChatGPT is a search engine

Regulators do not classify technologies. They classify functions, and on 31 August 2026 the European Commission decided that what ChatGPT does is search.

The designation

ChatGPT was designated a very large online search engine under the Digital Services Act — the first standalone AI service brought under the bloc's heaviest platform rules. Reddit (about 57.2 million monthly users in the EU) and Roblox (about 46.6 million) were designated at the same time.

The threshold is 45 million average monthly recipients in the EU. OpenAI reported roughly 159.1 million for ChatGPT search over the six months to 31 March 2026 — more than three times over.

The reasoning, which is the interesting part

The Commission called ChatGPT a hybrid service that qualifies as a search engine because it can engage with and respond to users' prompts and queries, including by searching the web.

Note what that argument does not depend on. Not the model, not the training data, not whether the answer is generated or retrieved. It depends on what the thing is for from the user's side: someone arrives with a question and leaves with an answer about the world. By that test the classification is hard to argue with, and it is deliberately built to survive the technology changing underneath it.

What now applies

Within four months of notification, a designated service must assess and mitigate systemic risks, submit to annual independent audits and act on their findings, share data with the Commission and national authorities and give vetted researchers access, offer a recommender option not based on profiling where that applies, and keep a public advertising repository where that applies.

The enforcement ceiling is 6% of global annual revenue. DSA enforcement has so far produced fines totalling around €870 million.

Why it matters beyond Europe

Two reasons. The first is mechanical: audit trails, risk assessments and researcher data access are expensive to build twice, so a company that builds them for the EU generally runs them everywhere.

The second is definitional, and longer-lived. Once an assistant is legally a search engine, the questions that were settled for search — what counts as a ranked result, what a source is owed, what a user must be told about why they were shown something — reopen for a system that answers in one paragraph and shows no list at all. That argument has barely started.