Google lets publishers block AI Overviews

Google will have to let publishers keep their pages out of AI Overviews and stop the content from training its models, after the Competition and Markets Authority imposed a new conduct rule. The decision hits the core of Google’s AI‑driven search experience and arrives as the company reshuffles its AI leadership.

The CMA announced that the rule gives publishers effective tools to prevent their material from powering AI features such as AI Overviews. The regulator phrased it as a “world first” and said the change puts news organisations in a stronger negotiating position. Google confirmed the change on the same day, saying it will let sites opt‑out of AI Mode and AI Overviews independently of normal search results.

The CMA ruling and its mechanics

The Competition and Markets Authority’s decision targets two specific uses of publisher content. First, it bars Google from surfacing the material in AI Overviews, the concise answers that appear at the top of search pages. Second, it blocks the use of the same content for fine‑tuning Google’s large language models. The rule does not affect a site’s visibility in standard organic rankings.

Publishers can now toggle a setting in Google Search Console that signals the company to exclude their URLs from AI‑generated snippets. The CMA’s press release emphasized that the tool is “effective,” but the agency did not disclose any enforcement penalties for non‑compliance. The ruling marks the first time a regulator has required a search engine to separate AI‑derived features from the traditional index.

Google’s response and internal shake‑up

Google’s public statement framed the opt‑out as a “flexible” option for site owners. The company said the change would be rolled out over the next few weeks and would apply to all properties that use the Search Console. No timeline for a full rollout was given, and the announcement omitted any comment on potential impact to ad revenue.

At the same time, Google disclosed a senior leadership change that could shape how the company implements the ruling. John Giannandrea, the long‑standing head of Google’s AI and search divisions, announced his departure. Giannandrea, who joined Google in 2010 and later added search to his remit, said he wanted more hands‑on time with technology. Jeff Dean, the current head of the Google Brain project, will assume Giannandrea’s AI responsibilities, while Ben Gomes, VP of search engineering, will become the new search chief.

The reshuffle consolidates the Google Brain team under a single umbrella, a move analysts see as an effort to streamline AI development after the company’s rapid rollout of Gemini and other models. Dean’s track record includes leading large‑scale deep‑learning infrastructure, which may influence how Google handles the CMA‑mandated opt‑out at a technical level.

Why publishers care about AI Overviews

AI Overviews have already reshaped click‑through patterns. Independent data shows that organic click‑through rates drop by about 61 % when an Overview appears, because users often accept the concise answer without scrolling. For newsrooms that rely on traffic for ad revenue, the loss of clicks translates directly into lower earnings.

Beyond traffic, the fine‑tuning clause raises intellectual‑property concerns. Publishers invest heavily in investigative reporting and original analysis. If Google harvests that content to improve its language models without compensation, the value of the original work erodes. The CMA’s rule therefore addresses both a revenue‑impact symptom and a deeper ownership question.

The opt‑out does not guarantee that a publisher’s brand will be invisible to AI. Google can still surface a site’s headline or a snippet that does not directly quote the article. Critics argue that the rule is a minimal concession that leaves the broader data‑harvesting pipeline untouched.

The broader AI search ecosystem and Reddit’s rise

Google is not the only player pulling content into AI pipelines. A recent analysis of 1,300 keywords across 26 industries found Reddit appearing in 82.7 % of top‑30 Google results, with an average position of 4.2. The study linked Reddit’s dominance to a $60 million‑per‑year partnership Google signed in 2024 to use Reddit data for AI training.

That partnership drove a 1,328 % visibility surge for Reddit, making it the second‑most visible site in Google US searches behind Wikipedia. Reddit also tops AI citation charts: it accounts for 46.7 % of Perplexity citations and 21 % of Google AI Overviews. Trust metrics reinforce the trend—90 % of consumers say they trust Reddit for product discovery, while trust in traditional reviews has fallen from 79 % to 42 %.

The confluence of AI Overviews, zero‑click searches, and Reddit’s data pipeline creates a feedback loop. When an AI model references a Reddit thread, the thread climbs in visibility, feeding more training data back to the model. Publishers that lack a Reddit presence risk being invisible not only to human users but also to the AI agents that now answer many queries.

For Google, the CMA ruling could complicate this loop. If a news site opts out of AI Overviews, the model may still draw on Reddit discussions that reference the same story. The separation between a publisher’s own domain and third‑party platforms becomes fuzzier, raising questions about the practical effectiveness of the opt‑out.

What to watch

The next few months will reveal whether Google’s technical implementation respects the CMA’s fine‑tuning ban. Track the rollout of the Search Console opt‑out toggle and any reported mismatches between publisher settings and actual AI behavior. Monitor Jeff Dean’s early decisions on model training pipelines—particularly whether Google shifts more data sourcing to public forums like Reddit. Finally, watch for follow‑up regulatory actions in the EU or US that could extend the opt‑out model beyond the UK. Online publishing’s new dynamics will be shaped by the interplay between legal constraints, leadership changes, and the growing power of AI‑driven search.