Brussels made a classification decision: chatbots that search for answers online are considered search engines. The platform label governs user-generated content, while the search engine label governs how queries are organized and presented. By placing ChatGPT in the latter category, it means that recommendations, citations, and retrieval paths must be reported based on systemic risk rather than just model safety.

The capital pathway is to front-load compliance costs. The 159 million monthly active users in the EU push OpenAI from the "innovation sandbox" onto the same oversight table as Google and Bing. The four-month window requires piling up materials for audits, protection of minors, and election risks. Money flows from subscriptions and APIs into legal and trust safety teams, with an additional evaluation gate before product features go live. Ireland continues to act as the coordinating country, with fines and procedures still following the very large tier of the Digital Services Act.

The benchmarks are Google Search and Bing being labeled as very large search engines, while TikTok and Meta are labeled as very large platforms. AI is at a stage where "conversational interfaces swallow search boxes": users no longer open a search engine first, but regulation extracts the search function separately from chat. The simultaneous entry of Reddit and Roblox into the platform tier indicates that the Commission is filling two nets on the same day: generative retrieval and user-generated worlds.

Structurally, this belongs to regulatory change. The mechanism is: as queries shift from ten blue links to a generated answer, risks expand from ranking bias to hallucinations, citations, and conversations with minors. The classification decision precedes technical terminology. It is permissible to call it a chatbot, but it is regulated as a search engine. Whoever crosses the 45 million threshold first enters the annual audit.

ABAB News · Cognitive Law

Conversations that can retrieve will be regulated as search engines.
Once user scale crosses the threshold, classification takes precedence over product terminology.
Generated answers replace blue links; the risk list will only grow longer, not shorter.