How Google AI Overviews chooses which sources to cite

You can rank number one on Google and still be invisible in the AI Overview above it. The signals are different, and most companies are solving the wrong problem.

Why are companies invisible in AI Overviews?

Organic click-through rate on US Google searches hit 40.3% in Q1 2025, the lowest ever recorded (Datos/SparkToro). For a growing share of queries, the answer now sits at the top of the page. Not a blue link. An AI Overview.

A company can hold position one and still be invisible in the AI answer above it. The signals that earn an AI Overview citation are different from the signals that earned the ranking. Traditional SEO gets a page indexed. It does not get it cited.

80% of informational queries now trigger AI Overviews (Semrush, 200,000 AIOs, 2025). Those are the queries buyers use when researching suppliers. They read the AI answer before they see any organic result, and the companies cited in that answer are not necessarily the ones ranking highest.

Google controls the organic index and the AI layer above it. Unlike ChatGPT (Bing's index) or Perplexity (its own index), Google's AI and organic search share the same index but apply different selection criteria. Ranking well does not mean Google's AI is citing you.

How does Google choose what appears in AI Overviews?

Google does not pull the top-ranking page. It runs query fan-out: splitting a query into sub-queries, retrieving results for each, then selecting pages that surface consistently across all of them. Entity confidence and structured data outweigh traditional ranking signals.

Google does not publish the complete AIO selection criteria. The patterns below are the strongest signals from independent research and from audit work in the field. The mechanics shift over time, which is what makes this a continuous discipline rather than a one-time fix. Four signals stand out.

Query fan-out

Google breaks complex queries into sub-queries and checks which pages surface consistently across the expanded set. Pages appearing most frequently become citation candidates. A page can rank well for one keyword and still miss the AI Overview for a broader query. Consistency across sub-queries matters more than ranking for any single one.

Entity confidence

Google's Knowledge Graph determines how clearly it understands what a business is, who it serves, and why it is authoritative. Named authors, organisational attribution, and topic-specific credentials create structural advantage. Most B2B companies have weak entity presence, which is the gap entity SEO closes by clarifying those brand, service, and audience relationships for Google.

Structured data

Schema markup tells Google's parser what a page is, who wrote it, and what it covers. Pages that require Google to infer purpose from raw prose get deprioritised versus pages with explicit schema declarations. Service, FAQPage, BreadcrumbList, and Person markup all contribute.

E-E-A-T as a selection filter

96% of AIO citations come from sources demonstrating real expertise, author credentials, and direct answers (Wellows, 2,400 citations, 2025). Domain authority alone does not get a page into the AI Overview. E-E-A-T signals do.

What is the difference between AI Overviews and AI Mode?

AI Overviews appear automatically in standard search results. AI Mode is a separate conversational interface inside Google Search. They share only 10.7% URL overlap (SE Ranking, 10,000 keywords, June 2025). Optimising for one does not guarantee the other.

AI Overviews are triggered by Google when a query benefits from a synthesised answer. The user does not opt in. They surface YouTube, Maps, and Google-owned properties preferentially, and they weight E-E-A-T and entity confidence heavily.

AI Mode is user-initiated. The user chooses to enter a conversational search experience, effectively ChatGPT inside Google. It draws on different sources, returns a different answer format, and follows different rules. 89% of URLs cited in one format do not appear in the other.

Treating "Google AI" as one optimisation target is a mistake. Which format buyers encounter depends on the keyword, so both need to be tested separately.

Source: SE Ranking, 10,000 keywords, June 2025

Google plays favourites. AI Overviews surface YouTube, Maps, and Google-owned properties preferentially. Businesses with YouTube presence and optimised Google Business Profiles benefit from a signal that does not exist on any other platform.

What does optimising for AI Overviews actually involve?

AI Overviews sit above organic results, and that changes the game. The work builds on SEO foundations but targets different selection criteria: entity confidence, content extractability, and the E-E-A-T signals Google weights above traditional ranking factors. It is not generic SEO, and it is not the same work that wins citations on ChatGPT or Perplexity.

Three areas determine whether Google's AI cites a page or ignores it.

• 01
  E-E-A-T is non-negotiable96% of AIO citations come from sources with strong E-E-A-T signals (Wellows, 2,400 citations, 2025). Author credentials, expertise demonstrations, real case data. Google already knows the site. The question is whether it trusts it enough for the AI answer.
• 02
  Content structured for extractionGoogle's AI runs expanded sub-queries to build each AI Overview, so content needs to answer the sub-questions, not just the main query. That means a clear H2/H3 hierarchy, direct answers at the top of each section, and FAQ sections that match the questions buyers actually type.
• 03
  Multi-modal signalsGoogle self-preferences YouTube and Maps in AI Overviews. Text plus images plus video plus structured data gets 317% higher selection versus text-only. Schema markup (Service, FAQPage, Person) helps extraction accuracy. This is Google rewarding its own ecosystem.

SEO is the foundation, and AIO builds on the same index. Domain authority, content quality, and technical health are all necessary but no longer sufficient. Fixing the page roles, schema, and entity structure that AI systems read is the work of a Visibility Rebuild. To appear in ChatGPT, Perplexity, or Claude as well, the signals overlap but each platform has distinct mechanics. The broader picture is covered across the full AI search visibility set of platforms.

How to assess where you stand in AI Overviews

Assessing AI Overview presence means testing whether Google's AI selects you or your competitors for the answers above organic results. The method is to run real buyer queries, record every response, score the citation rate, and benchmark against competitors. The output shows exactly who Google trusts for a given set of keywords.

An AIO-specific assessment looks across several layers: AI share of voice, content extractability, crawler access, brand entity coherence, third-party presence, technical foundation, and content quality. It tests which queries trigger AI Overviews, whether you appear, who is cited instead, and what structural signals competitors have that you lack.

Organic CTR in Q1 2025, the lowest on record. The companies visible in AI Overviews are capturing the buyers who never scroll to organic results. The question is whether Google cites you or your competitors.

Source: Datos/SparkToro, 2025

The businesses that establish AIO presence now build compounding authority. The ones that wait will be paying to displace them later.