Brussels Just Called ChatGPT a Search Engine. That One Word Triggers the EU's Toughest Rulebook

On August 31, the European Commission designated ChatGPT a Very Large Online Search Engine under the Digital Services Act — the same legal shelf as Google Search and Bing. It's the first time an AI chatbot has been dropped into that category, and OpenAI now has until January to prove that a product built to write poems and debug code can also survive being audited like a search engine.

By Gustavo Henrique · About the author

BRUSSELS · AUG 31-SEP 7 ~6 MIN READ

Updated on September 8, 2026 · Editorial update: source review, context and clarity.

159.1 million. That's the average number of people across the European Union who used ChatGPT's search feature every month over the six months ending March 31, 2026, according to OpenAI's own disclosure to Brussels. It's more than three times the 45-million-user threshold the EU's Digital Services Act uses to decide which online services are big enough to warrant its strictest oversight. On August 31, the European Commission acted on that number: it designated ChatGPT a Very Large Online Search Engine, or VLOSE, placing OpenAI's product in the same regulatory tier as Google Search and Microsoft Bing. Reddit and Roblox were designated Very Large Online Platforms in the same announcement, after reporting roughly 57.2 million and 46.6 million average monthly EU users respectively. None of the three had ever faced this level of scrutiny before.

Fast Facts

On August 31, the European Commission designated ChatGPT a Very Large Online Search Engine under the Digital Services Act — the same legal shelf as Google Search and Bing.
It's the first time an AI chatbot has been dropped into that category, and OpenAI now has until January to prove that a product built to write poems and debug code can also survive being audited like a search engine.
That's the average number of people across the European Union who used ChatGPT's search feature every month over the six months ending March 31, 2026, according to OpenAI's own disclosure to Brussels.

Why a chatbot counts as a "search engine" at all

The Commission's reasoning turns on a technicality that OpenAI itself can't really dispute: ChatGPT doesn't just generate text from what it already knows, it can also go out and retrieve information from the live web in response to a prompt. That web-search capability, layered on top of a conversational interface, is what the Commission calls a "hybrid service" — and under the DSA's definitions, a hybrid service that fetches and ranks information for users qualifies as an online search engine, full stop, regardless of how unfamiliar that label feels for something that answers in paragraphs instead of blue links. ChatGPT is now only the third service ever placed in the VLOSE category, alongside the two search giants it was implicitly built to compete with.

OpenAI, Reddit, and Roblox each declared surpassing the DSA's 45-million-user threshold themselves. The Commission didn't have to go looking for evidence that these services had grown large enough to matter — the companies handed it over.

What OpenAI actually has to do, and by when

The clock started the day the Commission notified OpenAI of the designation, and the company has four months from that point — by the Commission's own account, running to January 2027 — to meet a specific list of obligations that don't exist for smaller AI products. OpenAI will have to run and publish annual assessments of the systemic risks its search-enabled service creates, covering everything from the spread of illegal content to harm to minors' mental well-being to risks around elections and public security. Independent auditors, not OpenAI itself, will have to sign off on whether those risk-mitigation efforts are real. Vetted outside researchers get conditional access to OpenAI's data to check the company's homework. ChatGPT will need to offer at least one way of surfacing results that isn't built on profiling users' behavior, a public repository of any ads it runs, and a report on its content-moderation decisions every six months. Ireland's Coimisiún na Meán will supervise ChatGPT directly, working with the Commission, since OpenAI's EU entity is registered there; the Netherlands' Authority for Consumers and Markets does the same for Reddit and Roblox.

A second rulebook, not a replacement for the first

It's easy to miss in the headlines, but this designation runs on a completely separate legal track from the EU's AI Act, which already imposes its own transparency and risk rules on general-purpose AI models like the one powering ChatGPT. Legal analysts tracking the overlap have pointed out that OpenAI now effectively answers to two regulatory regimes at once, built around different definitions of harm and different reporting cycles, with real potential for the same product feature to trigger obligations under both laws simultaneously. That's not a hypothetical inconvenience — it means OpenAI's compliance, legal, and product teams now have to coordinate model testing and safety documentation across frameworks that weren't written with each other in mind, for a category of product that didn't really exist when either law was drafted.

How the companies involved are responding

OpenAI has been notably unbothered in public, with a spokesperson describing the company as already preparing to meet the DSA's additional requirements and framing the designation as a sign of how quickly Europeans have adopted AI-powered search — a company line that reads more like calculated calm than genuine surprise, given that Brussels had been signaling this exact move since a similar disclosure last October showed ChatGPT search already well past the threshold. Roblox took a similarly measured public tone: the company's European general manager, Joost Hagesteijn, said safety was central to how Roblox operates and that it looked forward to continued cooperation with the Commission. Reddit and the Commission itself have said comparatively little beyond the original announcement, leaving most of the public reaction to policy researchers rather than the companies now facing the deadline.

What comes next

January 2027 is the real test date, not August 31. That's when regulators find out whether OpenAI's first systemic risk assessment reads as a genuine accounting of ChatGPT's blind spots or as a compliance document written to survive an audit rather than to fix anything. Given that this is the first time Brussels has forced a generative AI chatbot through this specific process, expect regulators elsewhere — including in the U.S. and U.K. — to treat whatever OpenAI produces as the opening precedent for how AI search gets policed everywhere else.

This account draws on the European Commission's own press release, along with additional reporting from the European Commission, Tech Times, Tech Policy Press, and PYMNTS.

Gustavo Henrique Weighs In

The legal mechanics of this ruling matter less than the timing. Two years ago, “is ChatGPT a search engine” would have sounded like a philosophy-class question. Now it's a settled regulatory fact with an audit deadline attached. I don't think Brussels got the definition perfectly right — calling a conversational AI a “search engine” stretches a category built for ranked lists of links — but I also don't think they had a better option sitting on the shelf. The DSA was written for a web made of platforms and search bars, and generative AI didn't ask permission before becoming the way a few hundred million Europeans look things up. Rather than write an entirely new law, which would take years Brussels doesn't have, regulators reached for the closest existing box and made ChatGPT fit inside it. It's an imperfect fix, but an imperfect fix that arrives in January beats a perfect one that arrives in 2029, after the habit of asking a chatbot instead of a search engine has already become permanent.

A NOTE ON VERIFICATION

This story reflects our own reporting and verification process, not a reprint of any single outlet. Read more about that process in our editorial methodology.