Skip to content
50% OFF $299 $599
Lock in
§ 6.0.17 ARTICLE
Published VerifiedEvery 6 weeks Sources2 named Authored bySquareRank Team

Glossary · § 6.0.17 · Defined term

Passage ranking

Passage ranking is the search-indexing pattern Google rolled out in February 20212 where the engine ranks specific passages from a longer page rather than only the page as a whole — and the underlying retrieval pattern that AI engines (AI Overviews, ChatGPT, Perplexity) use to extract section-level answers. A page that ranks at #14 overall may still have one passage that ranks at #2 for the specific query that passage answers.

The 134-167 word self-contained passage rule that every cluster page on this site follows is essentially "make sure your sections are individually rankable as passages, not just collectively rankable as a page."

Definition

Passage ranking is the search-indexing pattern, rolled out by Google in February 2021, where the engine ranks specific passages from a longer page rather than only the page as a whole — and the underlying retrieval pattern that AI engines (AI Overviews, ChatGPT, Perplexity) use to extract section-level answers.

The unit of retrieval changed. Before passage ranking, Google's ranker scored pages as wholes — a single page-level score, possibly fragmented for display via featured snippets but ranked as one unit. After passage ranking, the ranker can score individual passages from a page independently and surface the strongest passage even if the page as a whole would not rank.

How passage ranking rolled out

Google announced passage indexing at its October 2020 'Search On' event<InlineCite n={1} sourceId='google-passage' />. The feature went live for English-language queries in the US in mid-February 2021. International and additional-language rollouts followed through 2021. The 2024-2026 AI Overviews and the broader AI-search-engine wave inherited the underlying retrieval logic.

The 2021 rollout had relatively quiet impact in classical SERPs — Google didn't dramatically reshuffle results overnight — but the change set up the retrieval pattern that all subsequent AI search products use. AI Overviews, ChatGPT with web search, Perplexity, and Claude all extract section-level rather than page-level when generating answers. The 2021 indexing change was the prelude; the 2024 AI-search wave is the consequence.

Implications for content structure

Three implications. (1) Long-form pages with clear section structure win — each H2 section is a passage candidate. (2) Sections should be self-contained — a reader (or engine) arriving at a single section should be able to understand it without reading the rest of the page. (3) Lead-first writing matters more than buried-the-lede storytelling — the first sentence of a passage is often what gets extracted.

The 134-167 word self-contained passage rule encodes this. Every H2 section on a pillar or cluster page opens with a bolded 1-2 sentence answer to its heading, then expands. The bolded lead is the extraction target; the expansion is the supporting context. The pattern shows up across this site because it follows directly from how passage ranking and AI retrieval work in 2026.

Passage ranking on Squarespace specifically

On Squarespace, passage ranking is editorial more than technical. The mechanical work is straightforward: use H2 headings consistently, open each section with a clear lead paragraph, keep sections roughly 200-500 words. Where Squarespace adds friction is the 7.1 section-based template emitting multiple H1s by default; passage ranking favours clean H1 → H2 → H3 hierarchy, and 7.1's default breaks that.

The Squarespace install pattern: audit each page's heading hierarchy in DevTools (the same audit we recommend for AI Overviews); Code-Inject a DOM rewrite to convert section-level H1s to H2s where the 7.1 template emits them incorrectly; rewrite blog-post intros so the first 200 words answer the post's primary query directly. The detailed cluster covers the install in step-by-step detail; see Squarespace AIO passages cluster.

Passage ranking is the technical companion to AEO and the input to AI Overviews retrieval.

Keep reading · § 6.0.17