The e-commerce figure sits close to the 20.5% all-SERP baseline. The SaaS figure is four and a half times higher than it.

The two sets differ in query shape and industry, and that is part of the same finding rather than a flaw. Transactional retail queries are short strings of product and brand names. The SaaS set is dominated by how-to and comparison phrasings. Industry and query shape are inseparable here because the industry determines the query shape.

What is AI Overview optimization?

AI Overview optimization is the practice of structuring a page’s passages so Google’s generated answer selects and cites them. This page uses that name throughout. The same discipline is also called Generative Engine Optimization (GEO) and answer engine optimization (AEO); the three refer to the same work at different levels of generality, and nothing here turns on which you use.

It operates at the passage level. Google extracts a definition, a list, or a direct paragraph rather than ranking the whole URL — which is the difference from classic SEO, where the unit is the page and the outcome is a position.

It sits alongside semantic SEO, the practice of building meaning through entities and relationships, and overlaps with LLM SEO, the work of surfacing in large language model answers.

AI Overview optimization matters because AI Overviews produce zero-click search results, where Google answers the query on the SERP and the user does not visit the source.

How do AI Overviews work?

An AI Overview works through a four-stage pipeline that turns indexed passages into a generated answer.

1. Google interprets the query using natural language processing (NLP), reading intent from the search string.

2. Query fan-out (splitting one query into related subqueries) generates related subqueries around the original intent.

3. AI Overview pulls passages from the index via passage indexing (extractable page sections such as paragraphs, lists, or definitions).

4. Gemini synthesizes the overlapping information into a generated answer with citations and link inclusion.

[PLACEHOLDER — W5 quote (a): Google’s sentence on what a page needs in order to appear in AI features, under 15 words, as a blockquote, with the line “Source: AI features in Search, Google Search Central” beneath it.]

AI Overview extracts short passages from indexed pages and combines matching claims from multiple sources into a single generated answer. The retrievable unit is the passage, not the whole page. The retrieval step itself is grounded rather than generated from memory; the mechanics are on RAG. So which signals decide which source the AI Overview selects?

What signals select an AI Overview source?

Google AI Overview selects a source using four signal groups: organic proximity, passage clarity, entity trust, and corroboration across sources.

• Organic proximity. AI Overview draws citations from pages that already rank in the top organic results.

• Passage clarity. Passage indexing (the scoring of individual passages within a page) determines which sections qualify. AI Overview extracts small self-contained sections such as direct-answer paragraphs, lists and definitions.

• Entity trust. Authority and trust signals tied to E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) and the Knowledge Graph carry weight. Examples are author identity, in-text citations, and structured data.

• Corroboration. Corroboration across sources is the most commonly repeated claim in this space — Respona states it as five sources agreeing — and it is also the one signal a SERP pull cannot isolate. Nothing here measures it.

[PLACEHOLDER — W5 quote (b): Google’s sentence on passage-level retrieval or query fan-out, under 15 words, as a blockquote, with the line “Source: AI features in Search, Google Search Central” beneath it.]

Retrieval reaches past the words the searcher typed, because Google splits the query first. The depth of that split per surface is covered on query fan-out.

Rich Sanger and Authoritas measured a 53% chance of citation for links in the first organic position, falling to 36.9% at the tenth. Authoritas and Optimizely separately put the share of AI Overview citations that also appear in the organic top 10 at between 40% and 76%. On the 91 AI Overview SERPs I pulled on 31 August 2026, the position-1 URL appeared in the citation set 71.1% of the time against 29.3% at position 8, with positions in between falling almost monotonically, and 47.2% of all citations pointed to pages that were not in the organic top 10 at all.

Surfer SEO measured citation share by domain. The general picture: