AI Search & GEO

How Google AI Overviews Choose Sources?

By BacklinkPrice Team · Nov 18, 2026 · Updated Nov 20, 2026 · 13 min read
Illustrated cover for How Google AI Overviews Choose Sources?

Google AI Overviews choose sources from indexed pages that answer the related sub-queries behind a question. Here is how fan-out, retrieval, and eligibility work, and how to get cited.

Google AI Overviews choose sources from pages that are indexed, eligible to show a snippet, and relevant to the related sub-queries behind a question. Google runs those sub-queries through its core ranking systems, then links to the pages that support each part of its answer.

Key takeaways

  • A page must be indexed and eligible to show a snippet in Google Search before it can appear as a supporting link in an AI Overview (Google Search Central, 2026).
  • Google says there are no additional technical requirements for AI Overviews, and no special markup is needed (Google Search Central, 2026).
  • AI Overviews may use query fan-out, which splits one question into several related searches across subtopics (Google Search Central, 2026).
  • Ahrefs found in March 2026 that only 37.9% of cited URLs also ranked in Google's top 10 for the same query, down from 76.1% in July 2025.
  • SE Ranking found in October 2026 that AI Overviews appeared for 83.2% of the queries it tracked, up from 28.66% in May 2025.

What AI Overviews are and how source choice works

An AI Overview is an AI-written summary that Google shows at the top of some search results, with links to the pages it drew on.

Google launched AI Overviews in the United States in May 2024, after testing the idea as Search Generative Experience in 2023 (SE Ranking, 2026).

The feature grew fast after launch. Google expanded it to more than 100 countries and territories in October 2024, when it said the feature reached more than 1 billion monthly users (SE Ranking, 2026).

By May 2025, Google had taken AI Overviews to more than 200 countries and territories and more than 40 languages, with 1.5 billion monthly users, rising to more than 2 billion by July 2025 (SE Ranking, 2026).

Source choice is not a separate ranking system. Google says AI Overviews use its core ranking and quality systems to find pages, and the model then picks supporting pages for the answer it writes (Google Search Central, 2026).

That point sets the entry test. A page that is not indexed, or that cannot show a snippet, is not in the pool of pages Google can link to in an AI Overview.

Why source choice matters and when it applies

Source choice matters because an AI Overview sits above the normal results and answers the question before a user scrolls.

SE Ranking tracked 100,000 keywords across 20 niches and found AI Overviews on 28.66% of queries in May 2025 (SE Ranking, 2026).

By October 2026, that share had reached 83.2% in the same tracking, with a peak of 84.16% in September 2026 (SE Ranking, 2026).

By September 2026, the gap had almost closed, with Business at 96%, Technology at 96%, Finance at 93%, Relationships at 93%, and Education at 92% (SE Ranking, 2026).

AI Overviews apply most to long, specific questions. SE Ranking found in May 2025 that 10-word queries triggered an AI Overview 69.21% of the time, against 25.82% for 4-word queries (SE Ranking, 2026).

That pattern fits informational intent. SE Ranking cites seoClarity research showing that over 96% of queries that trigger AI Overviews have informational intent (SE Ranking, 2026).

How Google chooses sources, step by step

Google chooses sources in a short chain: decide whether to show an overview, fan the question out, retrieve pages, write the answer, and link to supporting pages.

Step 1: Decide whether an overview adds value

Google first decides whether an AI Overview would add value beyond classic search results for that query.

Google says AI Overviews appear only when its systems judge them helpful beyond classic results, so they often do not show at all (Google Search Central, 2026).

Google may use query fan-out, which means it issues several related searches across subtopics and data sources to develop a response.

Google defines query fan-out as issuing multiple related searches across subtopics and data sources, while its models identify more supporting pages during generation (Google Search Central, 2026).

This step is why one page can win a citation without ranking for the exact query typed. It may rank well for one of the sub-queries instead.

Step 3: Retrieve pages through core ranking systems

Google retrieves pages through its core ranking systems, not through a separate AI index.

Google says AI Overviews and AI Mode rely on core Search ranking systems to retrieve relevant, up-to-date pages from the Search index (Google Search Central, 2026).

Google's model writes the summary from the retrieved pages and identifies supporting pages as it goes.

Google says that while responses are being generated, its models identify more supporting web pages, which lets it display a wider and more diverse set of helpful links than a classic web search (Google Search Central, 2026).

The link list is part of the answer, not an afterthought. SE Ranking found in February 2026 that an AI Overview cited 15.22 sources on average (SE Ranking, 2026).

Step 5: Apply the eligibility test

Google applies one clear eligibility test before a page can appear as a supporting link.

A page can appear as a supporting link in AI Overviews or AI Mode only if it is indexed and can show a snippet in Google Search, under the normal Search technical requirements (Google Search Central, 2026).

Google adds that there are no additional technical requirements, so no new file type, markup, or tag opens the door (Google Search Central, 2026).

Meeting the test does not guarantee a citation, because Google picks only a few supporting pages for each answer. ## What the data says about which pages get cited

The data shows a clear break from page-one rankings. A top 10 rank still helps, but it no longer decides most citations.

Ahrefs found in July 2025 that 76.1% of AI Overview citations came from pages ranking in Google's top 10 for the same query, in a study of 1.9 million citations (Ahrefs, 2025).

Ahrefs' March 2026 update, as reported by AEO Rankings in 2026, found that share had fallen to 37.9%, in a study of 863,000 keyword SERPs and about 4 million AI Overview URLs.

The rest split almost evenly in that March 2026 update: 31.2% of cited URLs ranked in positions 11 to 100, and 31.0% ranked beyond position 100 or did not rank for that query at all.

Other studies use different samples and methods, so their numbers differ. The direction is the same: most citations now come from outside the top 10.

The gap is wider for other AI tools. Ahrefs found in August 2025 that only 12% of links cited by ChatGPT, Gemini, and Copilot appeared in Google's top 10 for the same prompt, across 15,000 prompts (Ahrefs, 2025).

Perplexity was the outlier, with 28.6% of its cited URLs in the top 10, while about 80% of AI citations did not rank anywhere in Google for the original query (Ahrefs, 2025). That contrast is useful context for how ChatGPT chooses sources, because each assistant retrieves pages in its own way.

SignalWhat the data showsWhat to do about it
Indexing and snippet eligibilityRequired for any supporting linkFix crawl blocks and noindex tags first
Query fan-out coverageMost citations now come from outside the top 10Answer the sub-questions around your topic
Answer placementCited text sits near the top of pages in AI studiesPut the direct answer in the first 2 sentences
FreshnessRetrieval targets relevant, up-to-date pagesUpdate dates, prices, and facts on key pages
Source type mixVideo and social sources hold a large shareAdd video and community proof where it fits
Topical breadthAI Overviews now cover most niches above 60%Cover each subtopic on its own page

Which source types Google cites most

Google cites a wide mix of source types, and video and social pages hold a large share of citations.

YouTube was the most cited domain in SE Ranking's September 2026 data, with 16.15% of all citations, up from 10.71% in February 2026 (SE Ranking, 2026).

Social platforms made up 27.8% of all citations in September 2026, up from 12.15% in November 2025, and 65.81% of answers cited at least one social source (SE Ranking, 2026).

Ahrefs' March 2026 data shows the same video tilt. YouTube accounted for 5.6% of all AI Overview citations in that study, and 18.2% of the cited URLs that did not rank in the top 100 (AEO Rankings, 2026).

Brand mentions in AI search also matter here, because a brand that other sites mention often is easier for an AI answer to name and link to.

What to do to become a cited source

Becoming a cited source is a content and technical job, not a single trick. The steps below follow Google's own advice and the citation data.

Step 1: Confirm the page is indexed and can show a snippet

Check the page in Google Search Console with the URL Inspection tool, and confirm it shows as indexed.

Google says a page must be indexed and eligible to be shown in Google Search with a snippet to appear as a supporting link (Google Search Central, 2026).

Remove any nosnippet, max-snippet:0, or noindex tag from pages you want cited, unless you mean to block them.

Step 2: List the sub-questions behind your topic

Write down the 5 to 10 smaller questions a reader must answer before they can act on your main topic.

Google's fan-out process runs related searches across subtopics, so a page that answers only the head term misses most of the retrieval paths (Google Search Central, 2026).

Step 3: Put a quotable answer at the top of each section

Start each section with one sentence that answers its heading, then add the detail below it.

AI systems quote passages, not whole pages, so a clear answer in the first 2 sentences of a section is easier to lift than a slow build-up.

Step 4: Add facts, dates, and sources to every claim

Add at least one specific fact with a source and year to each section, and link to the original publisher.

Google lists helpful, reliable, people-first content as the core approach for AI features, and its best practice list includes text content that search systems can read (Google Search Central, 2026).

Step 5: Fix the technical basics Google lists

Google's best practice list for AI features is the same as its SEO basics: allow crawling in robots.txt, use internal links, give a good page experience, keep key content in text, and match structured data to visible text (Google Search Central, 2026).

Step 6: Build topical breadth and outside mentions

Publish a cluster of pages that covers each subtopic, rather than one long page that tries to cover them all.

Generative engine optimization works the same way: each page should answer one question well enough to be quoted on its own.

Tools, time, and cost

You need only a small set of tools to work on AI Overview citations, and most of the work is writing and checking, not buying software.

Google Search Console is the first tool to use. Google says sites in AI features are reported in its Performance report, within the Web search type (Google Search Central, 2026).

The Generative AI performance report in Search Console shows impressions by page, country, device, and date, and Google says it rolled the report out to all websites worldwide as of August 31, 2026.

That report shows impressions, which count how many times links to your site were shown in a generative AI feature on Google Search (Google Search Console Help, 2026).

Check pricing on the vendor's own pricing page before you buy, as of November 2026, because plans and prices change often.

How to measure results

Measuring AI Overview results needs 2 views: Google's own reports, and your own checks of which pages get cited.

Start in Search Console. Open the Generative AI performance report and sort by page to see which URLs earned the most impressions in AI Overviews and AI Mode (Google Search Console Help, 2026).

Then open the main Performance report, within the Web search type, to see clicks and average position for the same pages, because AI feature traffic is reported there (Google Search Central, 2026).

Track citations by hand as well. Run your 20 to 30 target queries in a private browser window, record which URLs the AI Overview cites, and note whether each cited URL ranks in the top 10 for that query.

Use how to check backlinks as part of that review, because links from other sites still support the authority signals that feed retrieval.

Common mistakes

  • Chasing only the head keyword: most citations now come from sub-queries, so a page that answers only the main term misses the fan-out paths Google runs.
  • Blocking the snippet by mistake: a nosnippet or max-snippet:0 tag stops Google from using your text, and a noindex tag removes the page from the pool entirely.
  • Adding AI-only markup or files: Google says there is no special schema or file for AI Overviews, so that work adds nothing.
  • Writing one long page for every subtopic: a single page rarely answers each sub-question well enough to be quoted, while a cluster gives Google a clear page per sub-query.
  • Burying the answer at the end: a section that builds slowly to its point gives the model no clean passage to quote.
  • Buying links to force a citation: buying links is against Google's guidelines, and it does not fix indexing, snippet eligibility, or weak answers.

Frequently Asked Questions

Do AI Overviews only cite pages that rank in the top 10?

No. Ahrefs found in March 2026 that only 37.9% of cited URLs also ranked in Google's top 10 for the same query, down from 76.1% in July 2025. The rest ranked in positions 11 to 100, or did not rank in the top 100. A top 10 rank still helps, but it no longer decides most citations.

What is query fan-out in Google AI Overviews?

Query fan-out is Google's process of splitting one question into several related searches across subtopics and data sources. Google runs those sub-queries, retrieves pages for each one, and then writes a summary with supporting links. A page can win a citation by ranking well for a sub-query, even if it never ranks for the exact query typed.

Do I need special markup to appear in AI Overviews?

No. Google says there are no additional technical requirements to appear as a supporting link in AI Overviews or AI Mode. A page must be indexed and eligible to show a snippet in Google Search. No special schema, file, or tag is needed, but any structured data you use must match the visible text on the page.

How many sources does an AI Overview cite?

The count has fallen over time. SE Ranking found an average of 15.22 sources per AI Overview in February 2026, then 7.84 in July 2026 and 7.66 in October 2026. AI Mode cited more, at 9.43 sources per answer in October 2026. Fewer sources per answer means each citation slot is harder to win.

Can I opt out of AI Overviews?

Yes, with the same controls used for search snippets. Google says you can use nosnippet, data-nosnippet, max-snippet, or noindex to limit what Google shows from your pages. Google-Extended is a separate control for AI training and grounding in some of Google's other systems, not the switch for AI Overviews in Search.

How do I track whether my pages appear in AI Overviews?

Use the Generative AI performance report in Google Search Console, which Google rolled out to all websites worldwide as of August 31, 2026. It shows impressions by page, country, device, and date for AI Overviews and AI Mode. Check the main Performance report too, because clicks from AI features are reported there within the Web search type.

Sources

If you want to compare what links cost once a page earns visibility, see our pricing page.

BacklinkPrice Team
Written by
BacklinkPrice Team
Link building and backlink pricing research

The BacklinkPrice Team researches backlink prices, publisher sites and link building tools. We check prices and plans against each provider's own pages and date them, so you know how current the numbers are.

More from BacklinkPrice Team

Related guides