EU orders Google to share Search data with rival engines and AI chatbots

The European Commission adopted a decision on 16 July 2026 specifying how Google must share Search data with rival search engines under Article 6(11) of the Digital Markets Act. Google has to hand over anonymised ranking, query, click and view data, and the list of eligible recipients explicitly includes AI chatbots that offer search functionality. This is not a finding that Google broke a law. It defines what compliance has to look like from here, with the first obligations landing in August 2026 and the last in January 2027.

What the Commission decided

The decision closes a specification proceeding opened in January 2026. A specification decision is a technical instrument rather than a penalty. Article 6(11) of the DMA already obliged Google to share search data on fair, reasonable and non-discriminatory terms. What was missing was any agreement on what sharing meant in practice. The Commission has now written that down on its Digital Markets Act developer portal, down to named data fields, eligibility tests, anonymisation thresholds, audit cycles and a pricing formula.

This is a separate matter from the self-preferencing case, which asks whether Google favoured its own services inside search results. Different question, different proceeding. This one is about who gets access to the data.

What data Google must share

The scope covers anonymised ranking, query, click and view data generated by end users. In practice:

• queries entered into Google Search
• query metadata, such as language and device type
• URLs viewed by users
• user interactions with search results
• ranking information

The exclusions matter as much as the inclusions. Google does not have to hand over user account information, search histories, precise timestamps, very long queries, queries containing rare words, location specifics, or the URLs of paid results. That is where the privacy and commercial lines were drawn, and the carve-outs take a lot of the sharpest signal off the table.

Who is eligible to receive it

Recipients have to be online search engines, and the Commission has said AI chatbots offering search functionalities count as such. An applicant must also meet each of these criteria:

• operate as a genuine economic activity for at least two years, or have been founded less than two years ago with at least 50 million euros of capital investment
• have a minimum of 50,000 monthly users in the EU
• not be under EU sanctions or controlled by high-risk third countries
• process the data inside the EEA, with GDPR-equivalent protections

Naming AI chatbots in the eligibility list is the part with the longest shadow. It settles, at least for this remedy, that an assistant answering questions is a search product.

How the anonymisation works

Direct identifiers such as usernames and IP addresses are removed. Rare terms and unusually long queries are suppressed. On top of that sits a k-anonymity requirement: each user must sit in a group with "at least 1,000 users with the same location, device type and query language", and the Commission expects 95% of users to fall into considerably larger groups of 29,000 or more.

Verification is not left to Google. An independent audit is required before access begins, another within six months of sharing starting, annual audits after that, and the Commission can order ad hoc audits whenever it wants.

What recipients are allowed to do with it

One thing, essentially. Recipients may use the data only to "develop and optimise their own search technology and services". Training general-purpose AI models is prohibited, so is improving non-search services or advertising, and so is systematically replicating Google's results. That last restriction is the one to watch, because the boundary between learning from ranking data and reproducing a ranking will be argued about by lawyers for years.

The deadlines and the price

Two parameters cap what the data is worth. Minimum latency is seven days after a query is entered, so nobody is getting a live feed. Maximum access duration is five years per beneficiary. And it is not free: Google may charge compensation reflecting "incremental costs incurred by Alphabet for sharing the search data and a reasonable return on capital". That return cannot exceed Alphabet's weighted average cost of capital, with limited exceptions capped at Google Search's operating margin.

Why this matters for search

Query, click and ranking data at Google's scale is the moat. A competitor can build a crawler and an index and still rank badly, because it has no record of what people actually asked for and what they clicked once they saw the options. Forcing access to that on FRAND terms is the most concrete structural remedy anyone has attempted against Google Search.

Whether it moves the market is a separate question. Seven-day latency, no rare queries, no paid URLs and a ban on replicating results are real handicaps, and a rival still has to pay for the privilege. But if even a couple of rival engines or AI assistants get measurably better on the back of this, the medium-term result is a more fragmented search market, which for anyone doing SEO means more surfaces to be visible on rather than one.

What this means for Thai marketers

Directly, nothing. The remedy is EU-scoped and the Commission's pages make no claim about Thailand at all. Thai SERP behaviour does not change because of this decision, and there is no action item here for a Q3 plan.

The honest read is second-order and slow. Products built in the EU do not stay in the EU. If mandated data access makes rival engines and AI assistants better at answering questions, the search-diversification pressure Thai brands already feel from AI answers gets reinforced from a new direction. That argues for treating AI SEO and generative engine optimisation as standing work rather than an experiment, while the compounding effort still sits in organic search fundamentals that every engine reads the same way. Track this one. Do not restructure anything for it this quarter.

The short version

Google has been told what to share, with whom, at what latency and at what price. Nothing on a Thai SERP changes this year. What makes it worth watching is what a stronger set of rival answer engines does to the meaning of ranking two or three years out. If you want a read on where your brand stands today across Google and AI answer engines, our team is happy to look.