TL;DR: On July 16 the European Commission adopted its Digital Markets Act specification decisions against Alphabet, eleven days before the July 27 statutory deadline. Google must make anonymised search query, view, click and ranking data available to rival search engines on fair, reasonable and non-discriminatory terms — and the decision explicitly makes AI chatbots with search functionality eligible recipients. Permitted uses include grounding chatbot answers and improving retrieval and ranking. This is EU-scoped, licensed, audited and slow, so nothing in your client’s dashboard changes this quarter. But it undercuts the assumption most AEO strategy rests on: that answer engines are stuck retrieving from a weaker index than Google’s.
What was decided
The Commission published its decisions on July 16, covering two things at once: AI interoperability on Android, and the sharing of Google Search data. The search-data half is the one that matters for visibility work.
Google must share, in the Commission’s framing, the same data it collects to optimise its own search services — subject to anonymisation. Per PPC Land’s reporting on the specification proceedings, that means anonymised query, view, click and ranking data, at record level, daily, drawn from across Google’s search surfaces rather than one property. The Commission also set a pricing formula and an access process, so this is a licensing regime, not an open dataset.
What is not included matters just as much. Per Search Engine Journal, the decision excludes Google’s ranking algorithms and technology, account details, search histories, timestamps, and rare or lengthy queries. Recipients get behavioural signal, not the machine that interprets it.
The guardrails on recipients are strict. PPC Land lists them: no attempting to re-identify users or link datasets at record level, no sublicensing to third parties, no retention beyond 13 months, and no use outside optimising or improving search services. Recipients must pass assurance audits before first access and annually after, with access suspended for missed or qualified reports.
And then the line that should stop any AEO practitioner mid-scroll. Permitted uses, per the same reporting, explicitly include “grounding AI chatbot services, refining retrieval and ranking systems, web crawling and index building.”
Why this is an AEO story, not a competition-law story
We have argued here before that AI assistants mostly read someone else’s index — that the retrieval layer behind a chatbot answer is usually a conventional web index, and that this is why classic technical SEO still decides AI visibility. The practical consequence has been a comfortable one: whatever the answer engines lack, Google’s behavioural data has been theirs alone.
That is what the decision changes, in one jurisdiction, over several years. Click and ranking data is the feedback loop that separates a good index from a great one — it tells a retrieval system which of ten plausible pages actually satisfied the person who asked. Handing an anonymised version of that loop to engines that ground chatbot answers is a direct upgrade to the retrieval quality behind those answers for EU queries.
Three consequences worth planning around, none of them urgent:
1. The engine list gets longer, not shorter. The eligibility bar is low enough to matter. Search Engine Journal reports a threshold of 50,000+ monthly EU users plus security and investment requirements — that admits far more than the handful of engines anyone currently tracks. If retrieval quality stops being a moat, “which engines are worth measuring for this client” becomes a live question again rather than a settled one.
2. Your Google performance becomes an input to visibility elsewhere. If rival retrieval systems are tuned on anonymised Google click and ranking data, then the pages Google’s users reward are more likely to be the pages other engines retrieve. Google performance and AI visibility have been sold as parallel channels. In the EU, after this, they are partly coupled — which is an argument for keeping conventional technical SEO fully funded, not for treating AEO as a separate budget line.
3. The dates are not close. The machinery has to be built first: PPC Land reports Google must finalise the Search Dataset within three months of the final implementing act, with template licence agreements two months after it. On the Android side, Search Engine Journal cites obligations dated August 1, 2027 and August 1, 2028. Anyone selling you a strategy pivot for this in 2026 is selling you a pivot, not a strategy.
Google disputes both rulings. Kent Walker’s position, per Search Engine Journal, is that the measures risk undermining privacy and security guardrails for European users. Expect litigation, expect the timeline to move, and expect the eventual dataset to be narrower in practice than it reads on paper — that is how DMA remedies have gone so far.
What agencies should actually do
Very little, immediately — which is the honest answer and the one worth putting in a client email before someone else puts a panicked one there.
- Don’t rebuild reporting around a 2027 licence. Nothing about how your clients are cited today changes because of this decision.
- Do keep measuring per engine rather than in aggregate. If retrieval quality shifts unevenly across engines over the next 18 months, only per-engine citation measurement will show it. A single blended “AI visibility score” will absorb the change and tell you nothing.
- Do note EU/US divergence in your client’s tracking. This is a European remedy. If you serve clients on both sides of the Atlantic, the engines worth tracking may genuinely diverge by market, which is a reporting decision to make deliberately rather than discover later.
- Do resist the temptation to brief this as a Google-is-losing story. It is a data-access remedy with an audit regime attached. Clients who hear “Google is being broken up” from their agency and then read the actual coverage lose confidence in everything else in the deck.
Key takeaways
- The Commission adopted its DMA specification decisions on July 16, 2026, ahead of the July 27 statutory deadline, covering Google Search data sharing and Android AI interoperability.
- Rival search engines get anonymised query, view, click and ranking data at record level, daily, on FRAND terms with a set pricing formula — but not Google’s ranking algorithms, account details, search histories or timestamps.
- AI chatbots with search functionality are explicitly eligible recipients, and permitted uses include grounding chatbot answers and refining retrieval and ranking systems.
- Recipients are tightly constrained: no re-identification, no sublicensing, a 13-month retention cap, and annual assurance audits with suspension for failures.
- Timelines are long — dataset and licence templates within three and two months of the final implementing act, and Android obligations dated August 2027 and August 2028 per Search Engine Journal. Google is contesting the decisions.
- The strategic read: better retrieval data for answer engines couples conventional search performance more tightly to AI visibility, and argues for measuring each engine separately rather than as one blended score.
Preferium measures citations across all four AI engines separately for exactly this reason — when the engines start diverging, an average is the last place you notice.
