The Unit of Competition Isn’t Always a Page
Most SEO mental models assume Google is choosing between pages. You write a page, Google evaluates it against other pages targeting the same query, and the best page wins. Clean. Tidy. Wrong — at least for a non-trivial slice of queries.
Since Google confirmed passage-level indexing in late 2020, the system can rank a specific passage from within a page independently of whether the rest of that page is relevant to the query. The page still loads. The passage still exists in context. But Google’s decision to surface that URL can be driven by 200 words buried 3,000 words into a 6,000-word guide — not by the page as a whole.
This has structural implications most sites aren’t accounting for. The long, comprehensive pillar page — the SEO staple — turns out to be both an opportunity and a liability under this model, depending on how it’s built.
What Passage Indexing Actually Does (And Doesn’t Do)
Google’s own framing: passage indexing is a retrieval mechanism, not a ranking factor. The ranking signals applied to a passage are still largely the same signals applied to pages — E-E-A-T, link equity, entity associations. What changed is what Google selects as the relevant unit when deciding whether a URL answers a specific query.
Before passage indexing, a long page covering ten sub-topics was evaluated primarily on its dominant topical signal. If the page was clearly about Topic A and mentioned Topic G briefly, it was unlikely to rank for a Topic G query — even if its Topic G section was genuinely excellent.
After passage indexing, that same page can compete for Topic G queries on the strength of that section alone. Google confirmed this affected roughly 7% of searches in English at launch — not a majority, but 7% of Google’s query volume is a very large absolute number, and it skews toward long-tail, specific, often high-intent queries.
The queries where passage indexing tends to activate are the ones where a user wants a precise answer that typically lives inside a broader document: deep technical questions, niche how-to steps, specific comparisons buried inside a larger guide. These are exactly the queries where long-form content was previously underrewarded.
Why Most Long Pages Are Poorly Structured for This
Long pages written to satisfy crawlers rather than readers tend to share a specific failure pattern: prose flows continuously, transitions are smooth, but each section doesn’t stand alone. Every passage assumes the reader has read the previous one. Definitions get introduced once and reused without repetition. Context is assumed throughout.
That structure works fine for a document someone reads start-to-finish. It works poorly when a passage is being evaluated in isolation by a retrieval system — or by a user who lands mid-page from a SERP.
The passage that ranks needs to be self-contained enough to answer the query without relying on surrounding context. In practice that means:
- The key term being targeted appears in that passage, not just in a heading two sections earlier.
- The passage contains enough specificity to be a complete answer — not a pointer to the answer somewhere else on the page.
- The passage doesn’t open with a sentence that requires prior context to parse. “As we mentioned above…” is a passage-indexing killer.
Each H2 or H3 section should be writable as a standalone answer, even if it’s also part of a larger flow. That’s a real structural constraint that changes how you plan long-form content from the outline stage, not the editing stage.
The Duplicate Targeting Risk
Here’s the edge case most people miss. If passage indexing lets a section of your pillar page rank for Query X, and you also have a standalone page targeting Query X, you may be creating a cannibalization problem that doesn’t show up in standard keyword-to-URL mapping.
Standard cannibalization analysis looks at which URLs Google is choosing between for a given query. But if one of those competing URLs is actually a passage within a pillar — not a distinct URL — the diagnosis is harder to spot. GSC will show the pillar page getting impressions for Query X and your standalone page getting impressions for Query X, and it looks like normal cannibalization. The fix isn’t the same, though. You can’t canonicalize or consolidate a passage out of existence without removing that content from the pillar entirely.
We’ve seen this exact pattern surface during cannibalization audits: a pillar page’s subsection quietly competing with a dedicated page for the same query. The fix is usually one of two things — either pull the overlapping sub-section from the pillar and let the standalone page own the query, or drop the standalone page and make the pillar’s passage stronger and more self-contained. Which depends on which URL carries more link equity and which query has more volume. No universal answer.
How to Actually Audit This
The audit is straightforward but requires manual work.
Start in GSC. Pull the queries where your long pillar pages are getting impressions but low CTR. Low CTR against decent impression share often means Google is surfacing the URL but users aren’t clicking — which can mean the snippet is pulling from a vague or incomplete passage rather than your strongest content.
For each flagged query, do the actual search and look at which part of the page Google is pulling into the snippet. If that snippet text corresponds to a passage that’s currently thin or incomplete, that’s a direct content edit target — not a metadata problem.
Then check whether that same query has a standalone page ranking or getting impressions. If yes, decide which page should own it and act on that decision. Don’t try to run both.
One more check: run a site: search on Google with the query appended and see which URL surfaces. If Google is consistently surfacing the pillar over your dedicated page, that’s a signal the passage is currently seen as more relevant. Either invest further in that passage or strengthen the standalone page’s signals until Google flips its preference.
Where Passage Indexing Intersects With AI Overviews
This matters more now because AI Overviews pull from passage-level content. The retrieval step that feeds an AI Overview is conceptually similar to passage indexing — Google is identifying specific passages within pages that answer the query, not always evaluating pages holistically.
If your long content is structured so each section stands alone, you’re simultaneously improving passage indexing coverage and expanding the surface area available for AI Overview citation. Those two benefits compound. A well-structured H3 section that opens with a clear claim, supports it with specific detail, and closes with a concrete implication is more likely to be pulled in either context than a flowing narrative that assumes prior context.
This is one place where the old advice — write for humans, not crawlers — needs a small update. Write for humans who dropped into this section with no prior context. That’s closer to how retrieval systems encounter your content, and it’s also a better reader experience for anyone landing mid-page anyway.
One Caveat Worth Keeping
Passage indexing is not a substitute for page-level authority. A poorly-linked domain with thin trust signals won’t outrank an authoritative page purely through better passage structure. The retrieval mechanism helps you compete more efficiently with the authority and link equity you already have — it doesn’t shortcut building them.
If a page has no external links and sits on a domain Google doesn’t trust yet, better passage structure will improve snippet quality at the margins but won’t move rankings meaningfully. Fix the authority problems first, then optimize passage structure to extract maximum value from whatever authority exists.
Worth checking before you do anything else: which of your current long-form pages are getting GSC impressions for queries you’d normally assign to a different standalone page? That’s probably where passage indexing is already active on your site — and where the clearest quick wins are sitting.