Definition
Passage ranking is a search engine's ability to judge a page by how well one specific passage on it answers a query, rather than only by the page as a whole. The name belongs to a Google ranking system announced in October 2020 and launched for US English queries on February 10, 2021. The same principle, scoring passages instead of pages, now sits at the center of how AI search engines pick their sources.
Google originally called the feature 'passage indexing' and then corrected the name, because nothing about indexing changed. As Google's Danny Sullivan explained at launch, Google still indexes whole pages and considers information about entire pages, but can now 'also consider passages from pages as an additional ranking factor'. The goal was to help long pages, or pages covering several topics, rank for a specific question answered somewhere deep inside them. Google said the change would improve about 7% of search queries once rolled out across languages; the US English rollout finished on February 15, 2021 (Search Engine Roundtable). It is distinct from featured snippets: passage ranking affects which pages rank, and the results look the same as any other result.
In AI search, passage-level scoring goes further. Answer engines split pages into chunks, retrieve and score the chunks individually, and hand only the best few to the language model, a process covered under sub-document retrieval. A reranking model then reads each passage against the exact query. And because systems such as Google AI Mode run many searches per question through query fan-out, a passage only has to be the best answer to one sub-query to be cited.
The effect on citations is measurable. In a March 2026 analysis of 863,000 keyword results pages and 4 million AI Overview citations, Ahrefs found that only 37.9% of cited URLs ranked in the top 10 for the original query; 31.2% sat in positions 11 to 100, and 31.0% did not appear in the top 100 at all. Page-level rank for the head term explains less and less of who gets cited.
Writing for passage ranking and for AI retrieval comes down to the same habits. Each section should answer one question, open with the answer, and carry enough context to stand alone: name the product or subject instead of saying 'it', and keep a figure next to the claim it supports. Headings should say what the section covers. Long pages are fine, and often better, when they are built from independent sections, because each section is another chance to be the best passage for some query. Pages written as a single continuous argument offer few passages that work alone.
When a page is fetched by AI crawlers but rarely cited, look at its passages before its authority. A competitor with a weaker domain but a clearer paragraph on the exact sub-question will usually win that citation. Answer-ready content and content chunking describe the editorial side of the same idea.
Examples of Passage Ranking
- A 5,000-word guide to home insulation ranks for 'can you insulate a cathedral ceiling from the inside' because one section answers that exact question, even though the page targets broader insulation queries.
- A buyer's guide section comparing two espresso machines' boiler sizes and warm-up times is cited in AI answers for comparison prompts, while the rest of the guide is never quoted.
- A small accounting blog is cited in AI Overviews for a narrow question on depreciation schedules, ahead of large finance publishers that only cover the topic in general terms.
- An SEO team splits a long FAQ answer into separate, clearly headed sections and sees several of them start appearing as cited sources for different fan-out sub-queries.
