How a Chatbot Ended Up Looking Like a ‘Very Large’ Search Engine
Under the EU Digital Services Act, a “very large online search engine” is any search service with over 45 million average monthly users in the bloc. OpenAI’s own DSA transparency report disclosed that ChatGPT’s search feature has 120.4 million monthly active users in the EU over a recent six‑month period, easily passing that line. That single data point has pushed the European Commission to consider classifying ChatGPT’s search mode as a major search engine rather than just a clever chatbot. Pure text‑generation tools were not the original target of the law, which was drafted before conversational AI search became mainstream. But once ChatGPT starts pulling live web results, curating them and presenting them back to you, regulators argue it behaves less like a private assistant and more like Google or Bing – and should be regulated accordingly under AI regulation Europe rules.
What VLOSE Status Would Actually Demand from ChatGPT
If ChatGPT’s search component is designated a Very Large Online Search Engine, OpenAI would face a tougher rulebook layered on top of existing AI law. The DSA would require systemic risk assessments about how conversational AI search might amplify disinformation, illegal content or other harms, and concrete mitigation plans. OpenAI would have to publish more detailed transparency reports explaining how its algorithms select and rank information, where content comes from, and how conversational responses are shaped. Independent audits and structured data access for regulators would become part of routine oversight. Stronger content moderation and accountability rules would also kick in, forcing OpenAI to clarify notices, appeal options and safety systems when users encounter harmful or misleading material. In practice, the ChatGPT search engine would be treated much more like a regulated platform than an experimental AI toy, tightening OpenAI transparency rules across the service.
When Search Rules Meet Chat: Ranking, Creativity and Opinionated Answers
Search regulation assumes a list of links that can be visibly ranked. ChatGPT, by contrast, gives you a single conversational answer – often mixing explanation, synthesis and links in one paragraph. Applying search‑style rules to this hybrid model raises tricky questions. Regulators want to know how content is ranked, yet conversational AI search blends sources and scores behind the scenes. OpenAI may need to expose more of that logic, perhaps by clearly labelling which parts of an answer are drawn from search results versus model knowledge, or by offering alternative ranked options users can expand. There is also tension around creative, speculative or opinion‑like outputs: should they be treated like ranked search results, editorial content, or something in between? Stricter rules could nudge ChatGPT toward safer, more neutral responses, potentially narrowing the playful or exploratory aspects that many users enjoy today.
Why OpenAI Is Wary — And What It Means for Other AI Assistants
Being treated as a VLOSE is not a badge of honour for OpenAI; it is a regulatory burden. The company already faces obligations under the AI Act, but DSA classification would add platform‑style scrutiny, more audits, and higher legal exposure if its systems spread illegal content or systemic risks. That likely means more lawyers and compliance staff, slower feature rollouts, and less freedom to rapidly experiment with new modes of search or advertising. OpenAI is not alone in this spotlight: if ChatGPT’s search engine qualifies, similar logic could eventually be applied to other AI‑augmented search tools from traditional engines such as Google Search or Bing, as well as newer assistants that blend chat and web browsing. Europe is signalling that once AI systems reach massive scale and influence, regulators will treat them like critical online infrastructure rather than purely experimental technology.
What Users Might Actually Notice Inside ChatGPT
For everyday users, the most visible changes would likely appear in the interface and in how answers are framed. Expect clearer labelling when ChatGPT is using live web search versus its internal model, along with more explicit safety notices and options to report problematic replies. You may see richer citation panels, with clearer links or source lists for key claims, and short explanations of why certain sites or facts were surfaced – echoing ranking transparency requirements for traditional search engines. Content filters will probably tighten, so some edgy or ambiguous topics could trigger gentler, more constrained responses. New dashboards or help pages might explain systemic risks, moderation policies and user rights in plain language. While all of this could make ChatGPT feel slightly more formal and less experimental, it also promises more predictable behaviour and clearer accountability when things go wrong.