The European Fee adopted two binding selections that require Google to share anonymized Search knowledge with rival search engines like google and to open components of Android to competing AI assistants.
The search knowledge measures give eligible suppliers, together with AI chatbots with search capabilities, entry to anonymized question, click on, view, and results-position knowledge they’ll use to construct their very own retrieval and rating methods.
The Fee set out each selections beneath the Digital Markets Act, six months after opening the proceedings that produced them.
We lined the search knowledge proposal in April, when it was preliminary findings out for public session. The model adopted this month is closing.
What the Choice Requires
Google is required to share anonymized knowledge on rankings, queries, clicks, and views from each free and paid Search outcomes beneath truthful and non-discriminatory phrases. This consists of data reminiscent of search queries, metadata like language and machine kind, seen URLs, person interactions, and end result positions.
Nonetheless, it doesn’t embody Google’s rating algorithms. Sure delicate knowledge, like account particulars, search histories, timestamps, and uncommon or prolonged queries, are suppressed to guard people.
The Fee said that Google’s present data-sharing method has failed. The brand new resolution particulars what efficient sharing entails, overlaying who qualifies and the way knowledge is priced, with prices primarily based on restoration reasonably than open-market charges.
AI chatbots that qualify as on-line search engines like google beneath the DMA are eligible to make use of the info to enhance their methods, however not for coaching normal AI fashions or replicating Google’s outcomes. These necessities are binding beneath the DMA however don’t contain fines, in contrast to a separate DMA case associated to self-preferencing and ongoing antitrust circumstances in European courts.
Why the Knowledge Issues for AI Search
This resolution extends past search engines like google into AI responses as a result of it includes grounding. AI chatbots use latest net knowledge to make sure their solutions are correct, and the standard of that knowledge is dependent upon the search data behind it. A 2025 explainer on AI Mode defined that Google grounds its fashions with a system referred to as FastSearch, which depends by itself search rating indicators.
The choice doesn’t give this to rivals. It doesn’t require Google to share FastSearch or its search algorithms and expertise. As a substitute, Google should share anonymized knowledge on queries, clicks, views, and end result rankings that eligible events can use to develop their very own retrieval and rating methods, with grounding being one of many authorized makes use of.
In February, we recommended that this EU course of might have extra vital long-term implications for AI search than the US antitrust case as a result of whether or not Google’s search knowledge powers competing AI instruments impacts the complete system of AI responses, citations, and referrals. The choice marks the purpose the place that challenge begins to be addressed operationally. A chatbot with entry to intensive anonymized Google Search interplay knowledge begins with a special baseline than one with out such knowledge.
Who Can Truly Use It
Which corporations profit first is dependent upon who can already successfully make the most of the info, not simply on eligibility. All candidates will need to have no less than 50,000 month-to-month EU customers and go both a two-year working historical past or, for newer entrants, an funding check. Safety screening and an impartial audit are required earlier than Google shares any knowledge.
Established search engines like google like Bing and DuckDuckGo are more likely to meet these thresholds shortly and act upon new knowledge rights. Newer entrants should develop the capability to leverage the info earlier than it influences their choices.
Within the brief time period, the impression on visitors can be restricted by a baseline we’ve tracked all year long. AI chatbots nonetheless account for a small portion of referrals. In line with SE Rating, all AI platforms mixed represented about 0.24% of worldwide web visitors as of January. Higher entry to this knowledge might affect what competing engines and chatbots can develop, but it surely alone doesn’t decide the place searchers go.
The Android AI Observe
The second resolution covers Android. Google should open a set of working system options to rival AI assistants so an individual can activate a competing assistant by voice, just like the “Hey Google” command, and let it act inside apps, reminiscent of reserving a taxi or drafting a reply.
Google should add most of those options within the subsequent main Android launch, Android 18, and by August 1, 2027 on the newest. Concurrent voice activation, which lets multiple assistant reply to totally different wake phrases, has a later deadline of August 1, 2028. Google’s personal Gemini assistant already has this degree of entry on Android, which is the asymmetry the choice is supposed to shut.
Google’s Response
Google disagrees with each rulings. Kent Walker, president of worldwide affairs at Google and Alphabet, wrote that they “threat undermining very important privateness and safety guardrails” for hundreds of thousands of Europeans. He additionally talked about that Google has repeatedly proposed options geared toward reaching the DMA’s goals. Relating to the search knowledge measures, his concern is about revealing European search knowledge to unfamiliar corporations with out correct anonymization or customers’ data and consent.
The Fee explains that the anonymization includes a multi-layered technical course of mixed with contractual safeguards, developed with inner and exterior privateness consultants. This course of permits Google to overview an applicant on cybersecurity or knowledge safety standards earlier than sharing any knowledge, and the measures might be reassessed if impartial testing reveals the safeguards are inadequate.
Why This Issues
As soon as suppliers go the entry course of, competing search engines like google and AI chatbots achieve entry to anonymized search knowledge, just like what Google has amassed at scale. A broader vary of suppliers utilizing this knowledge might allow extra search engines like google and chatbots that cite sources and generate referral visitors, shifting away from the present dominance of some platforms. Nonetheless, this doesn’t be certain that final result. It is dependent upon who qualifies by eligibility and audits, and the way properly the info performs in precise product growth.
Wanting Forward
Initially, searchers and publishers gained’t discover any adjustments. Google will spend the remainder of 2026 growing the dataset and establishing phrases, with its pricing proposal due by January 2027 on the newest. Every eligible supplier will then entry the info by itself schedule, after licensing it and agreeing to the value. The principle Android adjustments are due by August 1, 2027, and concurrent voice activation by August 1, 2028.
The Fee plans to overview these measures each two years and will reopen them if impartial testing signifies the anonymization is insufficient. Whether or not this can broaden the variety of engines and chatbots vying for visibility stays unsure, and the end result will solely be clear as soon as eligible suppliers start leveraging the info.
