Google's AI opt-out arrives without the click data

Google has begun letting publishers opt out of AI Mode and AI Overviews, but is not telling them what the cost of doing so would be. Search Engine Journal reports that the opt-out controls are rolling out without the corresponding click, impression or referral data segmented by AI surface. Publishers are being asked to choose between visibility and traffic without knowing the exchange rate.

The asymmetry is the point. Google has long bundled AI features into the same Googlebot crawl and the same Search Console reports that cover blue links. A site that blocks Google-Extended stops training data flowing to Gemini, but does nothing about AI Overviews, which run on the core search index. The new control finally separates AI Mode and AI Overviews from classic results. What it does not separate is the analytics. There is no row in Search Console that says "AI Overviews sent you 4,212 clicks last week, down 38% year on year." So opting out is a decision made in the dark.

Publishers have reason to suspect the dark is unflattering. Independent studies through 2024 and 2025, from Authoritas, Ahrefs and others, have shown click-through rates on queries that trigger AI Overviews falling by a third or more for the top organic result. Google's own line is that AI features send "higher quality" clicks. Both can be true. Neither is measurable from inside Search Console today.

The opt-out is a loyalty test, not a choice

Frame it commercially. Google is asking publishers to declare a preference (be cited by the AI, or not) while withholding the only number that would let them price the trade. A bank's investor-relations team weighing whether its quarterly commentary should feed AI Mode answers has no way to know whether the resulting citations drive any measurable traffic, or whether suppressing them would simply hand the citation to a competitor. A multilateral's policy brief faces the same problem at higher stakes: the audience for "what does the IMF say about sovereign debt restructuring" is increasingly an LLM, not a search results page, and the cost of invisibility there compounds.

The rational move, for most enterprise publishers, is to stay in. Not because the economics are good, but because they are unknown, and the downside of being absent from the answer a CFO or a minister reads is larger than the downside of forgone clicks the publisher cannot see anyway. Google knows this. The opt-out exists to satisfy regulators and the louder publisher trade bodies; it is not designed to be used.

What this changes for brands competing for citations

Three implications follow for B2B communications teams.

First, the citation itself is now the asset, not the click. If a brand cannot measure AI-driven referrals and cannot meaningfully withdraw from the system, then being named, quoted and linked inside the AI answer is the only outcome that can be optimised. That reframes the content brief. The page has to be structured to be extracted (clear claims, attributable data, named authors) rather than structured to win a click (curiosity gaps, gated assets).

Second, the measurement vacuum will be filled by third parties before Google fills it. Profound, Peec, Daydream and a growing list of monitoring tools already track LLM citation share across ChatGPT, Perplexity, Gemini and Claude. Expect procurement requests from CMO offices at industrial groups and financial institutions to start specifying "AI citation share" alongside share of voice within the next two budget cycles. Brands that wait for Google to ship the dashboard will be a year behind.

Third, the opt-out asymmetry is a template. Other model operators (OpenAI, Anthropic, Perplexity) will follow the same pattern: nominal publisher controls, opaque analytics, a presumption of inclusion. Any communications strategy built on the assumption that brands can negotiate AI presence on a per-platform basis is going to age badly. The negotiation has already happened, in Google's favour, and the only lever left is the quality of the content the models choose to cite.

The clean read on this week's announcement is that Google has done the minimum to be able to say it offers a choice. The real choice, for publishers and the brands that publish through them, is whether to keep optimising for a search results page that is shrinking, or for an answer they cannot yet measure but can still influence. The second job pays better.