Google AI Mode is the tab that answers your question conversationally instead of returning ten links, and it has moved from an experiment to something with a billion monthly users faster than any change to search in the last decade. If you run a website in the UK, it is already affecting your numbers whether or not you have looked.

What makes it different from an AI Overview is not just presentation. An overview sits above a normal results page that still exists underneath it. AI Mode replaces the results page entirely with a conversation, and the mechanism it uses to build an answer changes which pages get seen at all.

The short version: AI Mode takes one question, silently fans it out into many simultaneous searches, and assembles an answer from what comes back. You are no longer competing for a position on a query. You are competing to be the best source for one of a dozen sub-questions the user never typed, and most of those sub-questions are more specific than anything in your keyword research.

Query Fan-Out, and Why It Breaks Keyword Thinking

When someone asks AI Mode a broad question, the system does not run that query. It decomposes it. A question about choosing a payment provider becomes parallel searches about transaction fees, settlement times, integration effort, chargeback handling and regional availability, all executed at once, with the results synthesised into a single response.

The consequence is uncomfortable for anyone who has built a content strategy around head terms. The head term is no longer the unit of competition. The sub-questions are, and you did not choose them. Google did, based on what it inferred the person actually needed.

This rewards a specific shape of content. A page that thoroughly answers one narrow question well is more likely to be pulled into an answer than a page that covers ten questions shallowly, because the fan-out is looking for a confident source on each fragment rather than an overview of the whole topic. Comprehensive guides still have a role, but the citation tends to go to whoever answered the fragment cleanly.

It also means the traffic you receive arrives with unusual context. A visitor who clicks through from AI Mode has already read a synthesis, formed an initial view, and clicked because they wanted your specific contribution. They arrive further along than a normal search visitor and behave differently, which distorts your bounce and time-on-page numbers in ways that look like a problem and are not.

What Google AI Mode Growth Actually Looks Like

There is a lot of estimation in this area and very little first-party evidence, so here is ours.

Google Search Console added a Search Generative AI performance report that separates AI Overviews, AI Mode and Discover AI from ordinary results. It rolled out to UK properties first. On this site, that report shows generative impressions averaging around a hundred and twenty a day in the middle of May 2026 and around five hundred and fifty a day by the middle of August, with several days above seven hundred. That is roughly a fourfold to fivefold increase in three months on a site whose publishing rate did not change materially over the same period.

Two things follow from that. The first is that exposure inside AI features is growing much faster than conventional search exposure, so a flat overall impressions line can hide a substantial shift in where those impressions come from. The second is that the growth arrived without us targeting it, which suggests the driver is Google expanding the surface rather than anything a publisher did.

The click side is less encouraging and worth being honest about. Publishers in the UK have reported substantial traffic falls on queries where an AI answer appears, and our own experience matches the pattern: visibility inside AI features has risen far faster than clicks from them. Appearing in an answer is not the same as being visited, and any agency presenting generative impressions as a win without showing you the click column is selling you a vanity metric.

What UK Publishers Can Now Demand

Britain is the only major market where this is regulated rather than negotiated, and the rules are recent enough that most businesses have not registered them.

On 3 June 2026 the Competition and Markets Authority issued a final decision under Google’s strategic market status designation. Google must provide publishers with effective controls to prevent their content being used to power AI features in search, including AI Overviews, AI Mode and AI-generated features in Discover. It must attribute publisher content with clear links inside AI-generated results. And it must provide a separate opt-out covering the use of content for fine-tuning models. Google has nine months from that decision to comply in full, which places the deadline in early 2027, so parts of the tooling are still being built.

The stated intent was to give publishers leverage to negotiate payment for their content. Whether that materialises depends on scale, and for most businesses reading this the calculation is different. If your website exists to generate enquiries rather than advertising revenue, being cited inside an answer with a clear link is distribution you were not otherwise getting. Opting out removes you from the answer and protects a click you were probably losing regardless.

Media businesses with a subscription model face a genuinely different trade. Almost everyone else should be working out how to get cited, not how to disappear.

What To Actually Do About It

Start with whether AI systems can read you at all. Content that renders only in the browser, structured data that fails validation, and bot rules that block the crawlers behind AI features are all common, all invisible in ordinary rank tracking, and all disqualifying. Google documents what it expects in its guidance on AI features in Search , and it is worth reading rather than taking an agency’s word for. Our guide to how AI search engines read schema markup covers what these systems parse and where the parsing fails.

Then restructure for extraction. Every page that answers a question should answer it near the top, in a passage that stands on its own without the surrounding argument. That single change does more for citation rates than any amount of additional word count, and it costs nothing beyond editing.

Then move weight towards questions that cannot be answered without you. This is the durable part. If a question has a fixed public answer, the model already has it and will never need your page. If the answer depends on the reader’s own numbers, or comes from data only you hold, the model has to send someone somewhere. Interactive tools, original measurement and figures from your own operations are what survive the transition. Our generative engine optimisation guide sets out the reasoning in more detail, and optimising for AI Overviews covers the sibling surface.

Finally, change what you report. Rankings describe a results page that a growing share of your audience is no longer seeing. Generative impressions, clicks from those impressions, and branded search volume together give a truer picture than position tracking alone.

What This Does Not Mean

It does not mean conventional search is over. The great majority of commercial queries still return a normal results page, and transactional intent in particular is poorly served by a conversational answer. Somebody who wants to hire a developer in Manchester is not looking for a synthesis.

It also does not mean informational content is worthless. It means informational content has stopped being a reliable traffic source and has become a citation and credibility asset instead. That is a real change in what it is for, and budgets should follow it, but writing it off entirely gets the trade wrong.

Finding Out Where You Stand

The useful first step is a factual one: whether your pages are technically eligible for citation, and what your own generative report already says. Both have definite answers and neither requires a retainer to establish.

Mecanik runs technical SEO audits covering rendering, schema validity, entity structure and crawler access alongside conventional indexation checks, and handles the implementation through our website development team when a fix reaches into the application. If your traffic has softened over the last two quarters and the rank tracking cannot explain it, this is usually where the explanation is.

Related reading: AI SEO Agency: What They Do and What to Pay , SaaS SEO Agency: How to Choose One in 2026 , Optimize Your Website for ChatGPT Search and Perplexity and Website Migration Without Losing Traffic: 2026 Guide .

Frequently Asked Questions

What is Google AI Mode? AI Mode is a conversational search experience that replaces the conventional results page with a synthesised answer. It reached a billion monthly users by May 2026. Unlike an AI Overview, which sits above a normal results page, AI Mode substitutes for that page entirely.

What is query fan-out? Query fan-out is the technique AI Mode uses to answer a question. Rather than running the query the user typed, it decomposes the question into many more specific searches, runs them simultaneously, and assembles the answer from the results. Competition therefore happens at the level of sub-questions the user never typed.

Can UK websites opt out of Google AI Mode? The Competition and Markets Authority required Google on 3 June 2026 to give publishers effective controls to keep content out of AI Overviews, AI Mode and Discover AI features, alongside clear attribution and a separate fine-tuning opt-out. Google has nine months to comply. For businesses whose site generates enquiries rather than ad revenue, opting out usually removes distribution rather than protecting it.

How do I measure AI Mode traffic? Google Search Console includes a Search Generative AI performance report that separates AI Overviews, AI Mode and Discover AI from ordinary results, and it reached UK properties first. Read the impressions and clicks columns together, because visibility inside AI features has been growing far faster than clicks from them.

Does AI Mode make SEO obsolete? No, but it changes what is measured and what content is for. Most commercial and transactional queries still return a conventional results page. Informational content has shifted from being a traffic source to being a citation and credibility asset, which is a change in purpose rather than a loss of value.