Google's New Guide to Optimising for AI Search

The new resource is a direct response to two trends. First, AI Overviews and AI Mode are now a regular part of the Google results page, and they often determine whether a click ever lands on a publisher’s site. Second, the SEO community has spent the past year coining new acronyms (AEO, GEO, LLMO) and selling new playbooks for them. Google’s guide pushes back on much of that noise while clarifying what actually influences inclusion in AI features.

For UK businesses relying on organic search, the timing matters. Google’s own data in the Search Console Performance report now blends classic web links with AI feature links under the same “Web” search type, so the visibility you do or do not earn in an AI Overview shows up in the same impressions and clicks columns you already monitor. Treating AI features as a separate channel is a category error: they are part of the existing Search surface.

How it works

The guide describes two distinct AI surfaces inside Google Search. AI Overviews are the short summary boxes that appear above traditional results on selected queries, designed to give a fast take on a complicated topic and provide links for deeper reading. AI Mode is a separate, longer-form experience aimed at exploratory and comparison queries, where users want a reasoned response rather than a list of ten links. Both rely on the same core ranking and indexing systems as classic Search, and neither requires a separate optimisation track.

A central concept in the new guidance is “query fan-out”. When a user submits a question, Google’s AI model issues several related searches across subtopics and data sources before composing its response. The example Google uses is “how to fix a lawn that is full of weeds”, which fans out into “best herbicides for lawns”, “remove weeds without chemicals”, and “how to prevent weeds in lawn”. The model then pulls supporting links from across that broader set of results. A page that ranks well for a narrow related query can therefore be cited in an AI Overview answering a wider parent query, even if it never ranked on the original SERP.

The other technical phrase worth knowing is retrieval-augmented generation, or RAG. The AI model is not generating answers from training data alone, it is retrieving relevant pages from Google’s live index and using those pages to ground its response. This is why eligibility for AI features is identical to eligibility for classic Search: a page must be indexed and allowed to appear with a snippet. There is no separate AI sitemap, no extra schema, and no AI-specific markup.

What is actually new for SEOs

Most of the foundational advice is familiar: useful, original content, clean technical structure, valid structured data, readable HTML, and sensible internal linking. What is new is how Google has reframed those fundamentals for the AI context.

The guide singles out “non-commodity content” as the highest-leverage lever. Google’s example contrasts a commodity title (“7 Tips for First-Time Homebuyers”) with a non-commodity one (“Why We Waived the Inspection and Saved Money: A Look Inside the Sewer Line”). The second piece is built on first-hand experience that no other writer can replicate, so it is harder for a model to summarise away. Google is signalling that thin, summary-style content can be regenerated by the model itself and is therefore unlikely to be cited as a supporting link.

The guide also tackles AEO and GEO directly. “Answer engine optimisation” and “generative engine optimisation” are not, from Google’s perspective, new disciplines. They are SEO, which is the same conclusion we reach in our own comparison of AEO and SEO and in our guide to getting cited by LLMs. The guide includes a mythbusting section that pushes back on the idea of creating separate content for every fan-out query or stuffing pages with question-and-answer pairs to chase AI citations. Google warns that doing so primarily to manipulate AI rankings can fall foul of the scaled content abuse spam policy.

A short note on AI agents is included at the end. Agentic browsing tools that summarise pages and complete tasks on a user’s behalf are flagged as an emerging space, but Google’s controls remain the same ones publishers already use: robots.txt for crawling, and the nosnippet, data-nosnippet, max-snippet, and noindex directives for surface controls. Site owners do not need to add new files or markup to manage how AI agents handle their content.

What this means for UK agencies and site owners

Several practical implications follow from the new guide. The first is that we can stop pricing separate “AI optimisation” packages on top of SEO retainers. If a client is asking for AI Overviews and AI Mode visibility, they are asking for SEO, and the same audit, content, technical, and link-building work applies. Quoting an additional fee for “GEO” or “AEO” is hard to defend once a client points at this guide. If you would rather hand the whole thing over, that is what our answer engine optimisation service covers, priced as SEO rather than as an add-on. The model-visibility half of the same work sits with our generative engine optimisation service, quoted on the same basis.

The second is that content strategy needs to lean further into first-party data, original research, and lived experience. Generic explainers that summarise widely-available information are the easiest content for an AI model to replace with its own summary, which means they earn fewer citations and fewer clicks. UK agencies are in a strong position here because so much SEO content on UK topics is still written from a US perspective. A genuinely UK-rooted article (with UK pricing, UK regulation, UK suppliers, UK case studies) is harder for a generative model to flatten and easier for Google to use as a supporting source.

The third is that performance reporting can simplify. AI feature impressions and clicks are already included in the standard Search Console Performance report under the “Web” search type, so there is no need to plug in a third-party “AI visibility” tracker for monthly reports. For UK clients running Google Analytics alongside Search Console, the guide notes that clicks from search pages containing AI Overviews tend to be higher quality, with users spending more time on the destination page. That is worth flagging when an organic click count drops but engagement metrics do not.

Practical considerations

A few caveats are worth keeping in mind. Google’s position that no new schema is required for AI features specifically does not change the existing value of structured data for traditional rich results, especially FAQPage, Article, Product, and LocalBusiness markup. Pages with valid, visible-text-matching structured data still earn more SERP surface than pages without it.

It is also worth re-checking how preview controls are implemented. The guide repeats that nosnippet, data-nosnippet, max-snippet, and noindex remain the controls for limiting how content appears in AI features. UK publishers who experimented with blanket Google-Extended blocks last year may want to revisit that decision, as Google-Extended governs AI training and grounding in some of Google’s other systems, not the AI features inside Search itself. The two controls are separate, and conflating them is a common mistake.

Finally, the guide explicitly warns against creating thin, near-duplicate variations of existing pages to chase fan-out queries. This is timely for any UK agency tempted to spin up dozens of “AI versions” of existing service pages. The honest read is that broad, well-written hub pages will outperform a sprawl of thin variants in the AI context, exactly as they do in classic Search.

FAQ