On 31 August 2026 the European Commission designated ChatGPT a Very Large Online Search Engine (VLOSE) under the Digital Services Act — the first AI chatbot to carry the label, which it now shares with Google Search and Bing. The trigger was OpenAI's own report of about 159.1 million average monthly active recipients of ChatGPT search in the EU over the six months to 31 March 2026, more than three times the 45-million threshold. Reddit (57.2 million) and Roblox (46.6 million) were designated as very large platforms in the same batch. OpenAI has four months — to January 2027 — to comply with the systemic-risk, audit, data-access and ad-transparency obligations, with fines of up to 6% of global annual revenue for breaches.
The legal point is narrow: the designation turns on ChatGPT's live web-search function, not on it being an AI product. The practical point is wide: a regulator with subpoena-grade data access has just decided that an LLM answering questions with web results is a search engine, and search engines have obligations to the sites they index. Sources: the Commission's designation, Search Engine Journal's and Business Standard's reporting, and Commissioner Henna Virkkunen's statement that the three services "will now be held to a higher standard of scrutiny and accountability".
What OpenAI now has to do
| DSA obligation | What it means in practice |
|---|
| Identify, assess and mitigate systemic risks tied to the service and its algorithms | Documented risk assessments covering illegal content, minors, well-being, fundamental rights, elections and public security — for a system that generates answers, not just ranks them |
| Annual independent audits, with a written response to findings | Third parties get to test how ChatGPT search selects, weights and cites sources |
| Data sharing with the Commission and national authorities | Regulators can demand ranking and retrieval data |
| Vetted-researcher access for systemic-risk studies | Academics get a legal route to study citation patterns — the first public look at how an LLM search engine chooses sources |
| A recommender option not based on profiling | Where applicable, an answer mode that does not personalise on user history |
| Public advertisement repository | Any ads inside ChatGPT become searchable public records |
Designation is not a finding of wrongdoing; enforcement is a separate process. The Commission has levied about €870 million in DSA fines so far, including €550 million against AliExpress in July 2026, so the 6% ceiling is not decorative.
Why this matters more than the fine
For the last two years, generative engine optimisation — getting your pages cited in AI answers — has been reverse-engineering from the outside. Which sources get picked, how recency is weighted, whether citations favour large domains: all guesswork from sampling answers. Two DSA obligations change that. Vetted-researcher access means peer-reviewed studies of ChatGPT's retrieval and citation behaviour can be built on real data rather than scraped samples. Annual audits mean the selection logic gets described in documents whose summaries become public.
The second consequence is precedent. The Commission has said, in effect, that an AI system with live retrieval is a search intermediary. Perplexity, Google's AI Mode (already covered under Google Search's designation) and any future Claude or Gemini search product with 45 million EU users inherits the same obligations. The category now exists; it will fill.
What changes for publishers and site owners
Nothing in the DSA gives you a right to be cited. But three obligations hand site owners an argument they did not have last month.
Transparency about source selection. Search engines under the DSA must explain the main parameters of their ranking. Applied to ChatGPT search, that means a public description of what makes a page more likely to be retrieved and cited. Expect it to confirm what our most-cited pages already show in practice: answer-first openings, dated specifics, structured data, and a clear author entity.
Researcher access. Within a year there will be studies with titles like "Citation concentration in LLM search: evidence from ChatGPT". If they show a handful of domains capturing most citations — the pattern every third-party sample has found — that becomes a competition question, and the DSA's systemic-risk framework is built to ask it.
Ad transparency. Ads inside ChatGPT (see why ChatGPT added ads while Claude went the other way) become a public repository. Every publisher will be able to see who is paying to appear in answers and against which queries. That is a market map nobody has had for AI search.
The GEO checklist that survives regulation
Whatever the audits reveal, the tactics that hold up are the ones that make a page a good source for a system that must justify its choices. Lead with the answer in the first hundred words, with the number or date that makes it checkable. Put the entity that stands behind the page in structured data — the schema generator covers Organization, Article and FAQ, and the schema template pack has the fifteen page types we ship. Keep meta titles and descriptions honest; a regulator reading "main parameters of ranking" will not list clickbait, and the meta tags previewer shows you the snippet as an engine sees it. Date your claims and update the date when you update the claim. And write for extraction: short standalone paragraphs that quote cleanly, checked against the readability checker — an answer engine cannot cite a sentence that only makes sense with the previous three.
Quick answers
Does the designation ban or restrict ChatGPT in the EU?
No. It adds obligations — risk assessments, audits, data access, ad transparency — and is explicitly not a finding of wrongdoing. Enforcement, if any, is a separate process.
What is the threshold for a Very Large Online Search Engine?
45 million average monthly active recipients in the EU. OpenAI reported about 159.1 million for ChatGPT search over the six months to 31 March 2026, more than three times the line.
When does OpenAI have to comply?
Four months from notification, which puts the deadline in January 2027. Annual independent audits and the researcher-access regime follow from there.
What are the penalties for non-compliance?
Fines of up to 6% of global annual revenue. The Commission has issued about €870 million in DSA fines to date, including €550 million against AliExpress in July 2026.
Does this affect Perplexity or Gemini?
Any AI search service that crosses 45 million EU users can be designated the same way. Google's AI Mode sits inside Google Search, which is already a designated search engine. For the state of AI answer engines this summer, see our GPT-5.5 Instant default-model analysis.
The EU has decided that the thing citing your pages is a search engine. Treat it like one. The AI Growth Team Pack includes the GEO audit agent we run monthly against ChatGPT, Perplexity and Gemini; the January 2027 compliance deadline is the date to re-run it, because the answers you get after the first audit will not be the answers you get today. Every product mentioned is available at wowhow.cloud — pay once, ship forever.
Originally published at wowhow.cloud
Top comments (0)