Something Inc.LoginSchedule a free consultation
TECHNICAL SEO

Google's EU Search dataset is open. What actually changes.

Google published the eligibility, data scope and pricing model for the European Search Dataset Licensing Program in early September, following the European Commission's July 16 order. Most enterprise SEO teams do not qualify to license it. All of them should be watching who does, because the competitive shape of search infrastructure just moved.

TECHNICAL SEOEU DMA

On September 1 Google refreshed its Search Central page for the European Search Dataset Licensing Program with the implementation detail that had been missing since the European Commission adopted its July 16 measures. The page now spells out eligibility, the four data components on offer, the FRAND pricing test, and the anonymization and cross-border rules. It is the most concrete disclosure Google has made about the program since it began sharing data in March 2024 under the DMA gatekeeper regime, and it changes the competitive shape of who can build a serious European search product in the next twelve months.

For most enterprise SEO teams the immediate answer to 'do we care' is 'not directly'. You are almost certainly not eligible. But the vendors, competitors and search alternatives that are eligible will be building on top of a real slice of Google's EU search behavior by the end of the year, and that changes how citations, comparison surfaces and ad targeting behave in ways that will show up on your dashboards regardless of whether you touched the form. This is the piece to read once so it does not surprise you in a QBR.

4
anonymized data components in the license: ranking, query, click, view
50,000
monthly EU users required as one of the eligibility thresholds
2 years
the operating history threshold, with a capital-based alternative for younger applicants
€50M
minimum capital investment for an eligible applicant that was founded within the last two years
THE SHORT VERSIONGoogle published the concrete rules of its European Search dataset program on September 1, and the eligibility bar is set to admit real competitors (search engines and AI chatbots with EU operations) while excluding shell entrants. Enterprise SEO teams will not be signing licenses. They will be dealing with the second-order effects: better-tuned AI answers, sharper competitor SEO tooling, and a European search market with more than one credible ranking model in it by mid-2027.

What Google actually published this week

The program is the direct output of Article 6(11) of the DMA, which requires gatekeepers to share search ranking, query, click and view data with third-party online search engines on fair, reasonable and non-discriminatory terms. Google was designated a gatekeeper, the Commission adopted specifying measures on July 16, and the September 1 page is Google's compliance surface. Search Engine Roundtable posted the detail the day the page went live, and Search Engine Journal covered the eligibility gates the same day.

Four data components are on offer, all anonymized. Ranking data is the ordered result set Google returned for a query. Query data is the queries themselves at an aggregated level, with the personal identifiers stripped. Click data is aggregated click behavior on those results. View data is aggregated impressions on those results, which is the counterpart the industry has never had access to at Google's scale. The four together are effectively an anonymized slice of what an EU search engine's own log stream would look like if Google's log stream were the reference.

DATA COMPONENTWHAT IT CONTAINSWHY IT MATTERS FOR A LICENSEE
Ranking dataThe result sets Google returned for a query in the EEA, anonymizedA licensee can train and evaluate a ranker against Google's actual output, not a synthetic proxy
Query dataThe queries themselves at aggregated, anonymized levelA licensee sees what the EU market is actually asking, at real volume, which changes what to index and what to answer
Click dataAggregated click behavior on Google's result setsA licensee has a satisfaction signal for evaluating whether its own ranker produced a better or worse answer
View dataAggregated impressions on Google's result setsThe counterpart click-through-rate reference that no non-Google engine has had at scale; enables real CTR modeling

Pricing is limited to the incremental cost of making the data available, plus a specified rate of return. That is FRAND in the DMA sense, and it is what makes the program economically real: a small European search startup does not have to fundraise against Google's data cost, only its infrastructure cost. That change alone reshapes the cost structure of building a European search product from scratch.

Who qualifies, and who does not

The eligibility gates are designed to admit legitimate competitors and exclude data-broker attempts. Recipients have to be online search engines (which the program explicitly defines to include AI chatbots with search functionality) operating in the EU or EEA, with at least 50,000 monthly average users. They have to be either two years old or, if younger, backed by at least fifty million euros of capital investment. They cannot be controlled by non-EEA state actors and cannot be under EU sanctions. Data that leaves the EEA has to move under equivalent protection.

The interesting selection effect there is the AI chatbot inclusion. The Commission and Google both explicitly acknowledge that a chatbot with search functionality is a search engine for the purpose of this program. That is not a small drafting choice. It means Perplexity, You.com, Andi, Neeva-style entrants, and the European AI labs building answer surfaces are all in the applicant pool. It also means a licensed applicant can legally train and evaluate its answer surface against Google's actual EU ranking behavior in a way that was not on the menu six months ago.

01Real European search engines qualify, brokers do notThe two-year operating history plus 50,000 EU MAUs threshold cleanly excludes shell entrants trying to arb Google's data into a broker product, while keeping the door open for actual search alternatives with meaningful EU footprint.
02AI chatbots are treated as search engines here on purposeThe program covers 'online search engines including AI chatbots with search functionality', which is the acknowledgement that the answer surface is now inside the definition of search. Enterprise SEO teams should read that sentence twice.
03The alternative capital path is the growth-stage windowThe under-two-years, €50M-invested alternative is what lets a well-funded European AI search startup qualify without waiting two years to reach the MAU threshold. That will be the more common route in the first year of the program.

The parties who will not be signing licenses are equally telling. Individual publishers, brand SEO teams, agencies, competitive intelligence vendors and rank-tracker companies are all outside the eligible set. That means the dataset does not become a public commodity. It becomes a licensable input for a narrow class of search infrastructure players, and the downstream products of those players are how the rest of the market sees the effect.

What the four data components mean for enterprise SEO

Even if your team is not eligible, the outputs built on this data are going to reach your dashboards. Three product categories are about to sharpen materially in the EU, and every one of them touches enterprise SEO.

European AI search answers get more accurate on EU-specific queriesA licensee training against Google's actual EU query and click distribution will produce EU-specific answers that reflect EU search intent, not a US-centric proxy. If your company sells into EU markets, the answer engines in those markets will get better at surfacing the right vendor for regional queries. Whether that surfaces you or a local competitor is a content and entity-mirror question, not a distribution one.
Rank tracking against non-Google engines becomes defensibleRank trackers have always had thin, sampled coverage of Bing, DuckDuckGo and the AI answer engines. A market with real licensees producing high-quality European ranking outputs gives rank-tracker vendors a comparable second dataset to test against. Expect Q1 2027 releases from every major rank-tracker with 'now covering European AI search' in the changelog.
Competitive intelligence tools sharpen on EU-only queriesAhrefs, Semrush and their peers all buy or partner for a share of European search data. The licensees who now hold direct access will either license derived data downstream (permitted under FRAND, but a separate legal review) or compete with the incumbents. Either path improves the resolution of the EU-only slice of the tools you already pay for.

The measurement discipline this argues for is the same one we walked through in the AI search measurement error bar framework. More sources of ranking data across a fragmented set of engines does not automatically produce a better picture; it produces a more diverse picture that has to be reconciled. Enterprise teams that already run a citation panel (see the AI answer surface readiness playbook for the shape) are the ones that will absorb the new data cleanly. Teams that outsource measurement to a single vendor dashboard are the ones who will find the numbers stop reconciling with each other in Q1.

Three second-order effects to plan for in Q4

The first-order effect (who can license the data) is settled. The second-order effects are what enterprise teams should be planning for in the next planning cycle, because they hit the work you already do.

EFFECTTIMELINEWHAT TO DO ABOUT IT
EU AI answer engines get materially better at EU intentQ4 2026 through Q1 2027, as first licensees ship trained productsAdd at least two EU-specific engines to your citation panel if you sell into the EEA; run local-language prompts, not translations
Rank trackers add second-source EU coverageQ1 2027 for the first credible releasesDo not switch vendors early; wait for the second-source coverage to prove it moves in the same direction as your own panel
EU-specific competitor tools improve resolutionRolling through 2027Ask your existing SEO tool vendor which licensees they source EU data from, or whether they intend to license directly; the answer is a real due-diligence question in your renewal
European regulators expect gatekeepers to expand scopeContinuous, through the DMA review cycleDo not build a strategy on the assumption that non-EU search data is next; the DMA's scope is explicitly the EU, and other jurisdictions are moving on different timelines

The last row is where a lot of vendor pitches over-reach in the next two quarters. The DMA is a European regime, applied to gatekeeper platforms serving the European market. It does not extend to the United States, the United Kingdom outside the DMA-equivalent territories, or the rest of the world. If a rank-tracker vendor tells you their new dataset covers 'global AI search accuracy comparable to Google', ask them where the data came from. If the answer involves a European license, the coverage claim is about the EEA, not the world, and the pricing should reflect that.

The audit action items for enterprise teams this month

Two concrete moves for the next thirty days, whether or not any of this ends up on your quarterly report.

First, add three EU-specific answer surfaces to your citation panel if you sell into the EEA at any material scale. That means at least one European AI answer engine (Mistral's Le Chat is the current default, but the pool will widen fast as licensees ship), and at least one localized run of a global engine (a French-language ChatGPT prompt, a German-language Perplexity prompt) run against localized versions of your comparison prompts. If your panel today is US-English only and you sell into Europe, your measurement is blind on the surfaces most likely to shift first, and this is the quarter to close the gap. Our reporting and analytics work treats regional panels as a separate track from global ones for exactly this reason.

Second, run an EU-specific technical-SEO check on the pages that would be the load-bearing answers for European queries. Hreflang, canonical, currency and locale-specific product data, entity properties in Wikidata mapped to the correct EU entity (not the US parent). None of this is new work. It is table-stakes work that gets neglected on marketing sites whose primary market is the US, and it is what the licensed European engines will pull from first when they train their rankers. If your Wikidata property says your headquarters is in California and your EU rev is in local subsidiaries the entity graph does not link to, the paraphrase you get in a French AI answer will be worse than the one an EU-domiciled competitor gets, and it will be a self-inflicted wound.

The wider frame is that the search market in Europe has, quietly and by regulation, become a market with more than one credible ranker in it. That is a fifteen-year change in a fourteen-month window, and enterprise SEO teams that treat it as a compliance story miss it. Treat it as a market-structure story. Adjust the panel, tidy the entity graph, ask the vendor renewal question, and plan the B2B enterprise work on the assumption that the number of engines your buyer might consult in 2027 is materially larger than the number they consulted in 2025. That is the audit action item, and the rest is downstream of getting that assumption right.

Read the Search Central page. Note who applies. Track which engines get licensed. And then adjust your measurement and your entity mirror for the market you are actually going to be selling into next year, not the one you were selling into last.

See where you are cited today

A free snapshot audit of your rankings and AI citations before we ever talk.

JB
Josh BernsteinMANAGING PARTNER, SOMETHING INC.

Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.

Free consultation

Let us be the last SEO agency you ever work with

A 30 minute call and a free audit of your SEO and GEO position. You keep the findings either way.