Shadowfetch

Technology

Europe’s Google order turns AI search into a reader-choice test

Europe’s binding DMA measures for Google Search data and Android AI assistants could reshape how search, AI answers, and publisher discovery work for EU users.

Portrait of Celine MoreauBy Celine Moreau7 min read
Europe’s Google order turns AI search into a reader-choice test

The European Commission has turned one of the internet’s oldest power imbalances into a live test for the AI era: if Google Search benefits from search data at global scale, should rival search engines and search-enabled AI assistants get a regulated path to some of that data too?

Under two binding Digital Markets Act specification decisions finalized this week, Google must make two changes in Europe. First, it must give competing AI assistants effective access to key Android features that Google’s own services, including Gemini, can use. Second, it must share anonymized Google Search data with eligible third-party search engines, including AI chatbots that offer search functions. Search-data sharing is due to begin in January 2027, while Android AI-assistant changes start arriving with the next major Android release and no later than August 2027 for most covered features.

For readers, this is not just another Brussels-versus-Big-Tech item. It goes straight at how information gets found, ranked, summarized and routed before it ever reaches a publisher’s site. The Commission is effectively saying that search and mobile AI are becoming the same reader gateway, and that gateway should not be locked to one company’s data advantage.

The order is narrow in form but broad in consequence. It does not require Google to hand over its search algorithm, make raw personal search histories public, or create a new European search champion by itself. It sets a regulated route for certain competitors to access ranking, query, click and view data that Google collects in relation to free and paid search, after anonymization and other safeguards.

The Commission’s Google Search Q&A says Google has held a market share of more than 90 percent in Europe for decades, giving it “an unparalleled wealth of user data” that helps improve its search algorithm and reinforce its position. Regulators said Google’s earlier compliance approach was not effective, including because it removed between 90 and 100 percent of unique search queries from the dataset and excluded AI chatbots that provide search services from the possible beneficiary pool. After two years of talks, the Commission chose a specification process to define what compliance should look like.

That detail matters because modern search competition is no longer just a blue-link race. AI assistants need fresh retrieval systems to answer current questions, cite sources, and avoid confidently stale responses. The Commission explicitly says eligible AI chatbots with online search functions may receive shared data, while also barring recipients from using the data to train general-purpose AI models, build unrelated consumer-profiling or advertising services, or systematically clone Google’s results. In other words: the EU wants the data to improve search technology, not become a backdoor training buffet.

The privacy architecture is the hard part. The Commission says no user account information or search histories are shared; precise timestamps are not provided; very long queries and queries with rare words are suppressed; location data is generalized; interaction durations are aggregated into intervals; and URLs for paid results are removed. Eligible recipients must pass independent verification before receiving data, face another audit within six months of access, and then annual audits after that. The Commission also says Google can review applicants against objective criteria and may exclude sanctioned entities or companies controlled by a third country that poses serious and structural cybersecurity or data-protection risks.

Google says that is not enough. In a July 16 blog post, Kent Walker, Google and Alphabet’s president of global affairs, said the decisions “risk undermining vital privacy and security guardrails for millions of Europeans.” Google argues that exposing private searches to unfamiliar companies without adequate anonymization or user consent could weaken privacy, risk business trade secrets and endanger national security. On Android, Google says the ruling could grant external apps sensitive device permissions without the current safeguards provided by phone makers.

The Commission’s answer is that privacy and security remain baked into the framework, not waved away. Its press release says the data-sharing decision uses a multi-layer anonymization method developed with internal and external privacy experts and aligned with draft joint guidance on the DMA and GDPR from the Commission and the European Data Protection Board. It also says Google may assess whether sharing data with a specific third party would pose serious cybersecurity or data-protection risks before any data is handed over. The Commission can amend the decision later, including based on independent third-party evaluation and market developments.

The Android side is just as important for the next version of search. The Commission says about 60 percent of EU mobile users are on Android devices, making Android a central distribution channel for AI assistants. The decision covers 11 Android feature areas relevant to AI services, grouped around invocation, context, actions across apps and the operating system, and access to resources. In plain English, rival assistants should be able to do things users increasingly expect from Gemini: launch by voice, understand screen or app context with user permission, take actions in other apps, and use sufficient device resources to complete tasks.

The Commission gives practical examples: a user could ask a chosen assistant to book a taxi, draft and send an email through a preferred email app, add items to a shopping list, translate speech, surface a flight number from an email when it becomes relevant, or automate a food order. Google must implement most measures in Android 18 and by August 1, 2027 at the latest. Concurrent hotword detection, which would allow multiple services to be triggered by voice invocation, is slated for Android 19 and by August 1, 2028 at the latest.

That timeline is why this story belongs in the newsletter lane. Nothing changes for most users this morning. But the direction of travel is now clearer: the AI assistant is becoming the new front door to search, apps and media discovery, and regulators are trying to set the locks before one default service owns the hallway.

For publishers, the upside is not automatic. More credible search alternatives could mean more traffic sources and more competition in search advertising, as the Commission argues. It could also mean more answer engines summarizing publisher work before readers click through. The difference will depend on product design, attribution, citation quality and whether new search entrants treat the open web as a source ecosystem rather than a quarry.

The safest read is that Europe is making a bet on contestability. It is not declaring Google’s products bad or competitors good. It is saying that when a designated gatekeeper controls a dominant search engine and the dominant mobile operating-system channel for AI assistants, equal access rules become infrastructure policy. Google is right that search data is sensitive. The Commission is right that scale data is the moat. Both facts can be true, which is exactly why this decision is more consequential than a routine compliance order.

The next proof points are concrete: Google’s application forms and beneficiary information page by the end of August 2026; template license agreements and test samples by September; a finalized anonymized dataset by November; and pricing and communication to eligible providers by January 2027. If those milestones slip, or if privacy reviews show the anonymization is weaker than promised, the story changes. If they hold, Europe may become the first large test market for a more plural version of AI search.

Sources


Shadowfetch is a technology publication. Explore Shadowfetch Linux — our own Linux build — and the Shadowfetch apps on the App Store.

Sources

The piece is argued from European Commission specification decisions and Q&As, Google’s public response, and a cited technology news report, with the author drawing newsletter-style analysis from those materials.

Evidence types: opinion, regulatory documents, public statements, technology news report

Links verified

See a problem in this story? Report an error · Corrections policy · Our methodology

The Daily Download

One morning email: the day’s biggest technology stories — AI, new devices, and the companies shaping them.

Double opt-in. Unsubscribe anytime. See our privacy policy.