EU Labels ChatGPT a Search Engine Under the DSA

Brussels has decided what ChatGPT is, at least for legal purposes. On Monday the European Commission designated OpenAI’s chatbot a Very Large Online Search Engine under the Digital Services Act, the first time a generative AI product has been placed in that category. The same announcement pulled Reddit and Roblox into the strictest tier of the DSA as Very Large Online Platforms. All three now have until the end of November to comply with obligations that were written with Google, Meta and TikTok in mind.

The classification matters more than the label suggests. For three years, AI companies have argued that chatbots are a new kind of product that does not fit neatly into existing rules. The Commission has now said, in effect, that when a chatbot searches the web and returns answers, it is doing what a search engine does, and it will be regulated like one.

How a chatbot became a search engine

The DSA reserves its heaviest obligations for services that reach more than 45 million monthly users in the European Union, roughly one in ten of the bloc’s population. OpenAI disclosed that ChatGPT’s search function averaged about 159 million monthly active users in the EU over the six months to March 2026, more than three times the threshold. Reddit reported 57.2 million and Roblox roughly 48 million.

The Commission described ChatGPT as a “hybrid service” that qualifies as an online search engine because it can engage with and respond to users’ prompts and queries, including by searching the web. That reasoning is narrow on purpose. It hinges on the live web-search feature, not on the underlying language model. An AI assistant that only answered from its training data would not obviously fall under the same rule. One that fetches and ranks live results does.

Henna Virkkunen, the Commission’s executive vice-president for tech sovereignty, security and democracy, said the designations mean the three services “will now be held to a higher standard of scrutiny and accountability in the European Union, in line with their large impact on our citizens and society.” Supervision will run through the Commission working with Ireland’s Coimisiún na Meán for ChatGPT and Reddit, and the Netherlands’ Authority for Consumers and Markets for Roblox.

What OpenAI now has to do

The obligations are substantial. Designated services must carry out annual systemic risk assessments covering illegal content, the protection of minors, effects on users’ mental and physical wellbeing, fundamental rights, electoral processes and public security. They must submit to independent audits, give regulators and vetted researchers access to data, and explain how their recommendation and ranking systems work.

For a search engine built on a probabilistic model, some of those requirements will be awkward. Google can point to an index and a ranking algorithm. ChatGPT generates its answers, and the question of how to audit a system whose outputs are not fully deterministic is one the Commission has not yet had to answer in practice. Expect the first risk assessment, due within four months, to be studied closely by every other AI company with European users.

The penalties are not theoretical. Euronews reports that the DSA has already produced around €870 million in fines, including a record €550 million against AliExpress in July, €200 million against Temu in May and €120 million against X in December 2025. Non-compliance can cost up to six percent of global annual turnover. With the designations, the number of platforms and search engines in the DSA’s top tier rises to 28.

The political dimension

This will not be received quietly in Washington. The Trump administration has already described Digital Markets Act fines against Apple and Meta as a “novel form of economic extortion” and has threatened retaliation against countries whose digital rules target American companies. Three more US-owned services under direct Commission supervision is likely to sharpen that rhetoric, and OpenAI’s growing ties to the US government make it a more politically sensitive target than a social network.

For OpenAI the strategic question is whether to comply as a search engine, litigate the classification, or restructure its European product so that web search sits in a separate service. Each option has a cost. Complying means opening the system to auditors. Litigating means years of uncertainty. Splitting the product means a worse experience for 159 million users.

The more consequential effect may be on everyone else. Perplexity, Google’s AI Mode, Microsoft Copilot and every assistant with a browsing feature now has a precedent to reckon with. The Commission has drawn a line, and it runs through the web-search button. For more coverage of AI regulation and the law, visit Mylistingo.