ChatGPT is now a search engine in the EU: what that means for GEO and AI visibility

On 31 August 2026, the European Commission designated ChatGPT a Very Large Online Search Engine under the Digital Services Act. It is the first AI assistant ever placed in the DSA's strictest regulatory tier, alongside Google Search and Bing. Reddit and Roblox were designated Very Large Online Platforms in the same announcement.

The compliance burden here lands squarely on OpenAI. What brands get is something else: a regulator has now formally classified an AI assistant as a search engine, on the basis of what it does.

Why ChatGPT counts as a search engine

The classification rests on what ChatGPT does, not on what kind of product it is. The Commission treated it as a hybrid service that qualifies as an online search engine because it responds to users' prompts and queries, including by searching the web.

That reasoning is capability-based, which makes it portable. Nothing in it is specific to OpenAI, to chat interfaces, or to generative AI as a category, so any assistant with a live web-search function and enough declared scale meets the same test.

Designation follows from that scale. Services declaring at least 45 million average monthly EU users cross the threshold, and OpenAI's own DSA disclosure put ChatGPT search at approximately 159.1 million average monthly active recipients in the EU for the six months ending 31 March 2026. That is more than three times the bar.

Henna Virkkunen, the Commission's executive vice-president for tech sovereignty, security and democracy, said the three services will now be "held to a higher standard of scrutiny and accountability" in the EU.

Where ChatGPT now sits among designated services

The Commission maintains a public register of every designated service with its reported user figure. Placing ChatGPT into it produces a comparison that has not been available before, because third-party estimates of ChatGPT's search scale have been inconsistent and unverifiable for two years.

Source: European Commission register of designated VLOPs and VLOSEs, information updated 31 August 2026. Figures are as reported by each provider, without prejudice to the Commission's assessment of the methodology used.

ChatGPT is now the second-largest designated search engine in the EU by reported monthly users, roughly 40 million ahead of Bing.

There is a stronger claim available here and it is not safe to make. These are self-reported active recipients under a compliance definition, not query volumes, and a shopper using both Google and ChatGPT is counted in both rows. The figure establishes scale. It does not establish share of search.

What the designation actually requires

The obligations land on OpenAI and they run to a deadline of January 2027, four months from notification.

1. Systemic risk assessment and mitigation.

OpenAI has to identify, assess and mitigate systemic risks arising from the service and its algorithmic systems. The Commission's listed categories cover illegal content, negative effects on minors, users' physical and mental wellbeing, fundamental rights, electoral processes and public security.

2. Annual independent audit.

An external auditor has to review compliance at least once a year, and OpenAI has to respond to the auditor's recommendations.

3. Regulator and researcher data access.

OpenAI must share data with the Commission and national authorities, and open platform data to vetted researchers conducting qualifying systemic-risk research.

4. Recommender transparency and an ad repository.

Where recommender systems apply, there has to be an option not based on profiling, plus a public repository of advertising.

Non-compliance can reach 6% of global annual turnover. Two points worth holding onto: designation is not a finding of wrongdoing, and enforcement is a separate process. The Commission can revoke a designation if a service stays below 45 million EU users for a year.

What are the implications for brands

The practical value of the designation sits in what it settles about where this work belongs.

GEO, generative engine optimisation, is the work of getting a brand and its products discovered, correctly described and recommended inside AI-generated answers. It has been called several other things, and the inconsistency has been an obstacle to getting it funded.

Agentic commerce optimisation overlaps heavily with what is variously called GEO (generative engine optimisation), AEO (answer engine optimisation), AI SEO, and AI search optimisation. The distinctions are mostly emphasis: GEO names the technology producing the answer, AEO names the behaviour being optimised for, AI SEO borrows the vocabulary marketing teams already had, and agentic commerce optimisation names the case where the answer ends in a purchase. All describe the same job.

The more useful question is where that job sits. Search and AI answers are one discipline running on one set of signals, and increasingly on one interface. Amazon's main search bar now recognises when an input is a question and routes it to Alexa for Shopping rather than to a results page.

A shopper experiences a single act of discovery. Treating SEO and GEO as separate disciplines with separate teams and budgets reflects how organisations are structured rather than how the algorithms work or how customers discover.

Where the two genuinely diverge is the scoreboard. Classic SEO measures a position in a ranked list of links. GEO measures whether you are among the sources an answer is assembled from, which makes citation and accurate representation the target. Same discipline, same inputs, two ways of counting.

The designation leaves the method untouched and changes the standing. Where teams have met the objection that AI assistants sit outside search, there is now a regulator's classification resting on exactly that capability. The change is one of governance rather than of technique.

What brands should do today

Nothing on this list is new work. What the designation supplies is the mandate to fund it. Each item maps to one of the 5 C's of Agentic Commerce.

1. Consolidate AI assistants into search, in reporting and in the team. (Context)

If AI assistants sit in an innovation line item, move them into search. A regulator has now classified ChatGPT on its search capability, which is the cleanest available argument for merging the budget, the reporting and the ownership.

The practical version of this: one team, one set of inputs, two scoreboards. Anyone running GEO in isolation from technical and content SEO is duplicating the inputs and fragmenting the measurement.

2. Audit how you are represented, beyond whether you appear. (Correctness)

Designation raises formal scrutiny of algorithmic systems, and misrepresentation is a brand risk before it is a visibility one. Presence is the easy metric. Accuracy is the one that carries legal and commercial exposure.

3. Prioritise earned media and retailer listings. (Citations)

Azoma's analysis of millions of shopping agent citations found that in Q2 2026 ChatGPT's source mix ran at 41% earned media, 37% retailer, 19% UGC and 3% brand.com. Those first two are 78% of the citation base combined, and a brand invested only in its retailer listings is addressing barely over a third of what ChatGPT cites. This is the Citations C, and it is the one most brands underweight.

4. Fix contradictions between your sources. (Correctness)

When brand.com, retailer listings and third-party reviews disagree about a product, the model is the one that resolves the conflict. This is Correctness, and it is measured across sources rather than within your own. Whichever version is most consistently repeated across sources tends to win, whether or not it is the accurate one.

5. Baseline visibility before changing anything. (Completeness)

GEO tools, also sold as AI SEO platforms, exist to answer one question: what does the assistant actually say about your brand and your category today, and which sources is it drawing on to say it. The practical test of any GEO platform is whether it records the assistant's actual answer text rather than a rank position, whether it reports at product level rather than as a single brand score, and whether it identifies the citation sources behind each answer so you know where to intervene.

Prompt-level data is what makes the rest actionable, because ChatGPT answers the same question differently depending on phrasing. A brand can be named for one wording and absent from a near-identical one, so a single spot-check or an aggregate score identifies a problem without locating it.

➡️ Azoma tracks brand and product visibility across ChatGPT, Gemini, Alexa for Shopping and Walmart Sparky, with the exact prompts, answers and citation sources behind every mention, so you can see which sources are driving competitor recommendations and get included in them.

What this does not change

Three overclaims are already circulating. None of them survives contact with the actual text.

It does not give brands access to ChatGPT's ranking logic. The researcher access provision is for vetted researchers doing qualifying systemic-risk research, not for marketers or vendors. Risk reports and audits may become public in some form, and they are scoped to categories like minors and electoral integrity, not to why one moisturiser gets recommended over another. Expect useful context, not a ranking manual.

It does not change how recommendations are assembled. ChatGPT still builds shopping answers from cited sources, in roughly the mix above. Nothing in the DSA alters that, and a brand absent from the sources an answer draws on stays absent from the answer.

Paid placement remains separate. ChatGPT now carries sponsored ad units, and those sit apart from the organic answer rather than feeding it. Being named in a recommendation is earned through cited sources.

It is not a global change. The DSA is EU law and the designation covers the EU service. The capability-based reasoning gives other regulators a template they may copy, but a US or APAC team reading this as a worldwide regulatory shift is misreading it.

Important caveat on dates. Several outlets have reported a December 2026 compliance deadline. The Commission's own announcement says four months from notification, "i.e. by January 2027". Use the Commission's date.

Summing Up

For two years, the argument about whether AI assistants count as search has been a matter of opinion. As of 31 August, in the EU, it is a matter of law.

What brands get out of it is standing, and nothing more concrete than that. No new tooling, no ranking data, no placement to buy. The team that has been asking for AI search budget now has a regulator's classification behind the request, and the counter-argument has thinned out considerably.

The mechanics underneath have not moved. ChatGPT assembles answers from cited sources, 78% of them earned media and retailer listings and 3% of them your own site, and sponsored units sit outside that answer. The designation makes the work easier to fund. Getting cited still runs through the sources.

Azoma is built for exactly this. We measure how AI shopping agents represent brands and products across ChatGPT, Gemini, Alexa for Shopping and Walmart Sparky, and we fix the sources those answers are built from. Get in touch today for an assessment of where your brand currently sits in AI-generated answers.