The Maze: Europe is forcing Google to share part of the feedback loop that helped make its search engine dominant. From January 2027, eligible rival search engines and AI chatbots with search functions can request anonymised ranking, query, click and view data. The access is delayed, priced and ringfenced. It does not include Google's algorithm. Brussels is trying to turn a private data advantage into regulated infrastructure without turning private searches into public property.
The intervention targets Google's learning loop, not just its results page. The European Commission says Google Search has held more than 90% of Europe's market for decades. Scale produces queries, clicks and views; those signals improve ranking; better ranking attracts more users. The binding DMA measures give established engines, credible entrants and search-enabled AI chatbots access to a defined subset of that feedback. The intended beneficiaries can use it to improve query understanding, retrieval, ranking and indexing — the plumbing behind a credible answer.
Google's first compliance design made the dataset legal but nearly useless. The Commission says Alphabet's initial proposal removed 90–100% of unique queries and excluded AI chatbots providing search, resulting in no meaningful uptake. The final design tries to preserve more utility while lowering re-identification risk. Direct identifiers and linkable attributes are removed. Rare, long or sensitive-looking queries are suppressed. Location, device and language data is generalised so each user sits in a group of at least 1,000. The regulator says 95% will sit in groups of at least 29,000.
This is controlled access, not an open-data giveaway. Recipients must run genuine search services, show an investment plan and meet security and international-transfer requirements. The data must stay in a ringfenced environment. It cannot be linked with other datasets, disclosed onward or used to re-identify people. Independent verification is required before access, followed by another audit within six months and annual audits. Google can seek exclusions on public-security grounds. The Commission can reopen the decision if evidence shows the safeguards are failing.
The use licence draws a line between search improvement and model extraction. Recipients may use the dataset to understand queries, improve ranking and retrieval, decide what to index and ground current AI answers. They may not train a general-purpose AI model, build unrelated profiling or advertising services, or systematically reproduce Google's results. Data arrives with at least seven days of latency, and each beneficiary gets no more than five years of access. The PPC Land report captures the commercial tension: rivals receive learning signals, but not Google's algorithm, full internal telemetry or a perpetual mirror.
Price and usability will decide whether the remedy produces competition. Google may recover the incremental cost of providing access under fair, reasonable and non-discriminatory terms. That is more workable than free access with hidden rationing, but it creates a practical hurdle for small entrants. Google also argues the rules could expose private searches, trade secrets and national-security interests. The Commission's companion announcement says layered anonymisation, contracts and audits answer that risk. Neither side has outcome evidence yet.
Why it matters: Search and AI competition increasingly turns on feedback data, not just distribution. Europe is testing whether a regulator can prise open that advantage without handing over the algorithm or the user. If the dataset is useful and affordable, rivals could improve retrieval, merchants could gain more traffic sources and advertisers could face a less concentrated market. If suppression, price and compliance make access ceremonial, Google's flywheel stays intact. The policy is binding. The competitive result is still a hypothesis.
