Google AI Overviews: what gets cited and what you can influence

AI Overviews are the generated summaries Google shows at the top of some search results, with links to the pages they draw on. For many informational queries they are now the first thing a searcher reads. Google's newer AI Mode works on similar principles in a conversational interface. For site owners, two questions matter: how do pages get cited in these summaries, and what happens to traffic when they appear?

This guide sticks to what Google has said publicly and what can be measured. Much of what is written about "AI Overview optimisation" goes well beyond either.

What Google has said

• No special requirements. Google's guidance states that there are no extra technical requirements to appear in AI features. A page must be indexed and eligible to appear in Search with a snippet. There is no AI Overview markup, no special file, no submission form.
• Same systems, same principles. AI Overviews draw on Google's index and ranking systems. The guidance for appearing in them is the general guidance for Search: helpful, reliable, people-first content, and good technical foundations.
• Links are intended to be prominent. Google has described AI Overviews as a way to surface a wider range of sources, with links shown within and alongside the summary.
• Reporting is combined. Traffic from AI features is included in Search Console's Performance report, within web search totals, without a separate breakdown.

What decides whether a page is cited

Google has not published the selection process. Based on its statements and on what can be observed in results, these factors matter:

Factor	Why it matters	What to check
Indexing and snippet eligibility	Pages that are not indexed, or block snippets, cannot be used	No noindex; no nosnippet; not blocked in robots.txt
Ranking for related queries	Sources are drawn from pages Google already considers relevant	Visibility for the topic in Search Console
Clear, extractable statements	A summary needs passages that state facts plainly	Direct answers under clear headings; see featured snippets
Coverage of sub-questions	Overviews often combine several aspects of a question	Does the page address the follow-up questions?
Trust and experience	Google applies its quality systems to AI features too	Authorship, sources, first-hand detail; see E-E-A-T

A recurring observation is that Overviews often cite pages that are not in the top three organic results. A page that answers one specific aspect of the question especially clearly can be cited even from lower positions. That is good news for smaller sites with genuinely specific content.

Controls you have

Site owners have limited, blunt controls over AI features in Google Search:

• nosnippet (meta robots or X-Robots-Tag): prevents any snippet for the page, including use in AI Overviews. It also removes the normal text snippet from regular results, which usually costs clicks.
• data-nosnippet on an HTML element: keeps that part of the page out of snippets and AI features while the rest remains eligible. Useful for specific passages such as paywalled extracts or sensitive details.
• max-snippet: limits snippet length, which also limits how much can be used.
• noindex: removes the page from Search entirely.

Blocking Google-Extended in robots.txt controls whether content is used for training Gemini models and some grounding uses; Google has stated it does not affect inclusion in Search, and AI Overviews are part of Search. The AI crawlers guide explains the difference between search crawlers and training tokens.

Practical steps to be a better source

1. Fix the basics first. Pages must be indexable, fast and free of technical errors. A quick SEO check catches noindex, robots and canonical mistakes that would make a page ineligible.
2. Answer the question early and plainly. Put a clear, self-contained answer near the top of the relevant section. Overviews summarise; they need something unambiguous to summarise.
3. Cover the sub-questions. Look at the Overview and People Also Ask for your target queries. If the Overview addresses five aspects and your page covers two, there are three gaps.
4. Add what only you know. Specific numbers you measured, procedures you follow, examples from real work. Generic content is exactly what a summary replaces; specific content is what it cites.
5. Use structure that machines read well. Headings that match questions, real lists and tables, short paragraphs. The guide to content AI assistants quote covers this in more detail.
6. Keep information current. Prices, rules and product details that are out of date make a page a poor source.

Measuring the effect on your traffic

Because Search Console does not separate AI feature traffic, measurement has to be indirect:

• Identify queries that show an Overview by checking them manually or with a rank tracker that records SERP features.
• Compare CTR at similar positions for those queries before and after Overviews started appearing. A falling CTR at a stable position is the usual sign of answers being absorbed on the results page.
• Watch impressions: being cited in an Overview can add impressions for queries where you did not rank in the top ten.
• Look at engagement: visitors who click through from an Overview have often already read a summary and may arrive with a more specific need.

The Performance report guide explains how to compare periods properly. Tracking average position per query over time, as DeskSEO does when Search Console is connected, helps separate ranking changes from CTR changes.

Which queries are most affected

From observation, Overviews appear most often for informational and explanatory queries, multi-part questions and comparisons, and less often for navigational searches, very local searches and many simple transactional searches. Sites whose traffic depends heavily on short factual answers ("how many grams in a cup", definitions, simple conversions) tend to feel the effect most. Sites whose pages offer something you cannot get from a summary, such as tools, products, detailed walkthroughs, data, community discussion or first-hand reviews, tend to be affected less, and are more likely to be cited.

AI Mode and follow-up questions

AI Mode extends the same idea into a conversation: the searcher asks a question, reads a generated answer with links, then asks follow-ups. Google has described it as breaking a question into sub-topics and searching for each, which means a single conversation may draw on many more pages than a normal results page shows. The practical consequence is the same as for Overviews, only stronger. Pages that answer one specific sub-question very well have a chance to be cited, even if they do not rank for the broad query. A guide that covers the main question and its natural follow-ups, each with a clear answer under its own heading, gives the system more places to cite.

Clicks from AI Mode are also reported within Search Console's web totals, so the same indirect measurement applies.

What not to do

• Do not add hidden text or keyword lists in the hope of being picked up; Google's spam policies apply to AI features too.
• Do not rewrite every page into short question-and-answer fragments. Depth and context are what make a source trustworthy.
• Do not use nosnippet sitewide as a reaction without measuring; it removes your normal snippets as well and usually costs more clicks than it saves.
• Do not trust tools that claim to know Google's "AI Overview ranking factors". The system is not documented in that detail.

The longer view

AI features in search will keep changing in format and frequency. The pages that hold up are the ones that would deserve a click in any interface: accurate, specific, clearly written and maintained. Monitor which of your important queries show AI features, watch CTR and impressions on those queries, and keep investing in the content that a summary cannot replace. To see how your brand appears in standalone AI assistants as well, see measuring AI visibility.