How to optimise for Google AI Overviews and AI Mode: Google's official guidance [2026]

Google has published something SEO practitioners waited years for: official guidance on optimising sites for generative AI features in search. The document sits on developers.google.com and answers the question every site owner asks: what do I do to get my content into AI Overviews and AI Mode?

This article works through that guidance in detail, with practical advice, context from external research and steps to implement. The data is unambiguous: Semrush reports AI Overviews appearing on 88% of informational queries, and industry data puts AI Overviews on close to half of all Google results. If your site is not optimised for these requirements, you are losing visibility in the biggest algorithmic change in a decade.

How AI Overviews and AI Mode work: the RAG mechanism

To optimise for AI Overviews you first have to understand how these features work internally. Google describes two key mechanisms:

1. RAG - retrieval-augmented generation

RAG is the technique by which AI answers are not generated purely from training data but supported by real, current pages from Google's index. The system works like this: the user asks a question, Google runs its core ranking systems, retrieves the most relevant pages from the index, the AI model checks specific information from those pages, and generates an answer with visible links to sources. That means your page has to be well indexed and ranking for the relevant queries to enter the pool of sources AI draws on at all.

2. Query expansion

When a user enters a query, AI simultaneously generates a set of related, parallel queries that help gather more information. An example from Google's documentation: if someone asks "how do I clear a lawn full of weeds", AI may simultaneously search for "best lawn herbicides", "removing weeds without chemicals" and "how to prevent weeds appearing". For SEO that means content with broad semantic coverage - addressing a topic from multiple angles - has an advantage.

Does SEO still matter with AI in search?

Yes - and Google says so plainly in its guidance. But context matters. Ahrefs research found AI Overviews reduce CTR to pages ranking in top positions by 58%, up from 34.5% a year earlier. That does not mean SEO is dead; it means the goal has changed. It is no longer only about being in positions one to three in conventional results, but about being one of the handful of sources AI cites.

Semrush data suggests traffic from AI search may exceed traffic from conventional search by 2028. AI-cited pages already account for 17.31% of content in top Google results, up from 2.27% in 2019. That is a fundamental change in the search ecosystem.

Google's official priorities for AI Overviews

1. Valuable content that does not exist elsewhere

This is, according to Google itself, the most important factor - more important than every other piece of guidance combined. What Google specifically means by valuable content:

• A distinctive point of view - a first-hand review, an expert opinion grounded in personal experience, your own data gathered in the field. Not a summary of what is already online.
• Content that is not a commodity - Google's example: "7 tips for home buyers" is commodity content that anyone can write. "Why we skipped the survey and saved money" is a distinctive expert opinion.
• Clear structure - headings, paragraphs, a logical division of content. AI has to grasp quickly what the page is about.
• High-quality images and video - AI Overviews can display images and video from pages. Google encourages supplementing text with media.
• Not creating content for every conceivable query - producing hundreds of similar pages to manipulate ranking breaches the guidelines.

2. A clean technical structure

Technical SEO has not lost importance, but Google emphasises that the foundation is content accessibility to crawlers. The specific technical requirements:

• The page has to be indexed in Google Search and eligible to show a snippet. Without that, no AI will reach your content.
• Content has to be accessible to indexing crawlers - if you block GPTBot, Bingbot or Googlebot, you disappear from AI's field of view.
• Semantic HTML - it need not be perfect, but semantic structure helps screen readers and AI crawlers understand the content hierarchy.
• JavaScript SEO - Google can process JS, but working with it is harder. If key content loads through JS, make sure it is available to crawlers.
• Page quality for users - correct rendering on every device, low latency and a readable structure.
• Eliminating duplicates - duplicated content degrades the experience and can lead AI to select a different, original source.

3. E-E-A-T: experience, expertise, authoritativeness, trustworthiness

E-E-A-T is the framework Google uses to assess content quality, and it translates directly into AI visibility. Google's guidance points to specific signals:

• A clear author identity - who wrote the article, what experience they have, links to their profile. Anonymous content carries a weaker E-E-A-T signal.
• Transparent information about the site - who is behind it, how to make contact, the terms of use and privacy policy.
• Citations in external, independent sources - press, trade media, Wikipedia. That is an authority signal AI takes into account.
• Your own research and data - original statistics, case studies, study results. AI readily cites distinctive data it cannot find elsewhere.
• External reviews - Google Maps, Clutch, G2, Trustpilot. Independent validation of company quality.

The Ahrefs study of 75,000 brands showed brand mentions correlate with AI visibility three times more strongly than backlinks - 0.74 versus 0.27 for Domain Rating. The strongest signals are mentions on YouTube, Reddit and Wikipedia, platforms with high authority and a large volume of user-generated content.

4. Structured data (schema.org)

Schema.org markup helps search engines and AI systems understand the meaning of content on a page. Google does not say outright that schema guarantees presence in an AI Overview, but the data points to a correlation. The most valuable types:

5. Speed and Core Web Vitals

Google emphasises that high page quality for users, including load speed and Core Web Vitals, is part of the technical foundation for AI. Slow or unstable pages with high layout shift give worse experiences, which affects the quality assessment. Tools for checking: Google PageSpeed Insights, Search Console's Core Web Vitals report, and web.dev.

GEO versus SEO in practice

GEO describes optimisation aimed at being cited by generative AI systems, not only ranking in a conventional search engine. The differences in approach:

An important point: GEO does not replace SEO. The foundation is shared - an accessible, technically sound site with valuable content. GEO is an additional layer of work that improves the odds of being cited by AI.

How to check whether your page appears in an AI Overview

Google Search Console has no dedicated AI Overview report yet. The available diagnostic methods:

The manual method

1. 1Go to Google and search 10 to 20 category queries you want to appear on.
2. 2Check whether an AI Overview module appears above the results.
3. 3If it does, check whether your site is among the source links beneath the answer.
4. 4Repeat on different devices and in a private window, since AI Overviews render inconsistently.
5. 5Test Google AI Mode as well - the fully AI-integrated mode that represents the future of search.

The tooling method

• Semrush AI Overview Tracker - shows which keywords your site appears for in AI Overviews and how often.
• Ahrefs - has AI Overviews tracking in its SERP features section.
• SE Ranking - includes an AI Overview monitoring module.
• Rankscale.ai - a dedicated AI visibility tracking tool covering all models.
• Google Search Console - track changes in CTR and position. A sudden CTR drop at a stable position can mean an AI Overview is taking the clicks.

An implementation plan

Based on Google's official guidance and external research, a priority plan:

Phase 1 - technical foundations (weeks 1-2)

1. 1Check robots.txt - make sure you do not block Googlebot, GPTBot, Bingbot or ClaudeBot. Crawler access is a necessary condition.
2. 2Verify indexing in Search Console - check which pages are indexed and which have errors.
3. 3Add llms.txt - a file at the domain root describing the company, services and key links in an AI-friendly format.
4. 4Check Core Web Vitals - use PageSpeed Insights and fix critical performance issues.
5. 5Implement schema.org - at minimum Organization on the home page, Article on every blog post, and FAQPage on Q&A sections.

Phase 2 - content optimisation (weeks 3-6)

1. 1Audit existing content - which articles are commodity and which carry distinctive value? Focus updates on high-potential content.
2. 2Add data and statistics - original research, case studies with numbers, survey results. This is the hardest citability signal.
3. 3Expand author descriptions - who wrote the article, what experience they have, a link to LinkedIn or a profile. An E-E-A-T signal.
4. 4Add FAQ sections to key pages - questions and answers in a Q&A format with FAQPage schema are readily extracted by AI.
5. 5Structure content for query expansion - for each topic, cover multiple perspectives and related questions.

Phase 3 - building authority (months 2-4)

1. 1Earn mentions in industry media - guest articles, interviews, citations. Mentions outweigh backlinks for AI.
2. 2Build a presence on Reddit and Q&A platforms - answer category questions where AI learns.
3. 3Activate company profiles on Google Business Profile, Clutch and G2 - external quality validation.
4. 4Create original reports or research - proprietary data is the strongest magnet for AI citations.
5. 5Monitor and iterate - check AI citation rate, direct traffic and brand mentions monthly.

What NOT to do

Google's guidance warns explicitly against several practices that can harm you or produce nothing:

• Creating hundreds of pages for every conceivable query - that breaches Google's spam policy on scaled content abuse. A large page count does not improve site quality.
• AI-generated content with no expert input - AI can create content, but it has to be checked and enriched by an experienced person.
• Duplicating content - duplicated material reduces the chance of citation, because AI will choose the original.
• Blocking AI crawlers in robots.txt - if GPTBot is blocked, ChatGPT Search will not index your site.
• Ignoring technical work - without indexing the story ends. A page Google cannot see will not be cited by AI.
• Optimising purely for exact-match keywords - AI understands semantics. Keyword stuffing does not help and can harm.

Google AI Mode and why it changes things

AI Mode is a Google search mode in which the entire search experience is built on generative AI. Instead of a results list, the user receives an extended answer with analysis, sub-questions and cited sources. Semrush describes it as the future of Google Search.

The key differences from a standard AI Overview: AI Mode is used for complex, multi-step queries rather than only informational ones. Answers are longer and more elaborate. More sources are cited at once. Users can hold a conversation with follow-up questions. That means pages with deep, expert content have an even greater advantage in AI Mode than in standard AI Overviews.

Summary: Google's official guidance in ten points

1. 1Create distinctive, valuable content grounded in real experience - do not repeat what is everywhere.
2. 2Ensure technical accessibility - indexing, no crawler blocks, fast loading.
3. 3Implement schema.org structured data - Article, FAQ, Organization, HowTo.
4. 4Build E-E-A-T - author identity, external citations, original research.
5. 5Do not create content at scale purely for the algorithm - it breaches the guidelines and does not work.
6. 6Use semantic HTML and a correct heading structure.
7. 7Add llms.txt - define what your company is for AI.
8. 8Build brand mentions - more than backlinks.
9. 9Monitor AI Overviews in Search Console and external tools.
10. 10Focus on user satisfaction - AI algorithms are designed to connect users with content that genuinely answers their needs.

Google's official guidance confirms the direction we have observed for months: AI does not replace SEO, but it changes its priorities. Sites with distinctive, expert content, sound technical foundations and strong brand authority will be favoured increasingly - in conventional results and in AI Overviews alike.