How Google AI Overviews choose the pages they cite
AI Overviews are the generated summaries Google shows above some search results, with links to the pages they drew on. AI Mode is the conversational version. Google says there is no special trick to appear in either: the same pages that rank well and answer clearly are the ones that get cited. This article explains what that means in practice.
On this page — 5 sections
What are AI Overviews and AI Mode?
Quick answer
AI Overviews are short generated answers shown above Google results, with links to supporting pages. AI Mode is a separate, conversational search tab built on the same idea. Both launched in the US first and expanded in 2025 to many countries and languages, including Arabic.
Google launched AI Overviews in the US in May 2024 and announced at Google I/O 2025 that they had reached more than 200 countries and territories and over 40 languages, adding Arabic among others. AI Mode, a full conversational search experience, followed. For a clinic in Riyadh or a law firm in Cairo, this means some Arabic questions now get a generated answer before any blue link.
Each overview shows links to the pages it used. Those links are the new "position": a page can be cited even when it does not rank first for the query, because the overview is built from several pages that each answer part of the question.
How does Google pick the pages an AI Overview cites?
Quick answer
Google says AI features draw on its normal search index and ranking systems. A page must be indexed and eligible to show a snippet. Google also describes a "query fan-out" technique: it runs several related searches for sub-topics, then picks pages that answer each part well.
Google's Search Central guidance on AI features is short: there are no extra requirements and no special markup. To be eligible, a page must be indexed and eligible to be shown in Google Search with a snippet. The same ranking systems that decide normal results decide which pages are candidates.
Google has also said that AI Overviews and AI Mode may use query fan-out: issuing multiple related searches across sub-topics and data sources to build the answer. A question like "best way to cool a villa in Dubai in summer" may fan out to AC sizing, insulation, window film and electricity costs. A page that answers one of those sub-questions clearly can be cited for its part, even if it would never rank for the original question.
What makes a page more likely to be cited?
Quick answer
The same things that make it rank and satisfy: a clear, direct answer near the top, specific facts, original information others do not have, and content that matches the visible page. Google's 2025 guidance stresses unique, non-commodity content and a good page experience.
In its 2025 guidance on succeeding in AI experiences, Google listed a few priorities. They are familiar, but the emphasis is telling:
- Unique, non-commodity content: information people cannot get from ten other pages, like a Cairo movers company publishing its actual price ranges by apartment size.
- A clear answer up front: a heading phrased as the question, then the answer in the first sentence, so the passage can be lifted cleanly.
- Good page experience: fast, readable on mobile, with the main content easy to find.
- Structured data that matches the page: markup should describe what the visitor actually sees, never more.
- Images and video where they help: AI Mode can draw on multimodal content, so a real photo of the work beats stock images.
None of this is new. What changes is the reward: a page that answers one sub-question better than anyone else can now be cited even when it is not the overall winner for the query.
What controls do you have over AI Overviews?
Quick answer
The same snippet controls that apply to normal results. nosnippet, data-nosnippet and max-snippet limit what can be shown, and noindex removes the page entirely. Blocking Google-Extended does not remove you from AI Overviews; it covers Gemini model training and grounding.
Google documents that AI Overviews follow the standard preview controls. You can stop a page from being used with nosnippet, hide part of a page with data-nosnippet, or limit length with max-snippet. Each of these also affects normal snippets, so use them only where hiding the text is worth that cost.
| Control | What it does | Effect on normal results |
|---|---|---|
nosnippet | Prevents the page text being shown or used as a direct input in AI features | No text snippet in regular results either |
data-nosnippet | Hides a specific part of the page from snippets and AI features | That part never appears in snippets |
max-snippet:[n] | Limits how much text can be shown | Shorter snippets |
noindex | Removes the page from Google Search | The page disappears from results |
| Google-Extended (robots.txt) | Controls use for Gemini training and grounding | None: Search and AI Overviews are unaffected |
How do you measure traffic from AI Overviews?
Quick answer
Google counts clicks and impressions from AI features inside the normal Search Console performance report, without a separate filter. So you watch the pages and queries that matter, compare click-through rates over time, and check by hand whether your pages appear in overviews.
Google states that traffic from AI features is included in the Search Console performance report under the Web search type. There is no dedicated AI Overview filter. In practice, teams track a short list of important questions, search them by hand on a clean browser in the target country, and record whether their pages are cited.
This article is part of the Semantic SEO series — Writing for meaning: entities, definitions, attribute-value facts, semantic distance and pages that actually answer.
About the author
Mohamed Youns
Semantic SEO Engineer · Author & system developer
Mohamed Youns writes about how search engines understand content — the same standards he applies when building semantic systems at Nut Hub. nut-hub.org