Passage Indexing

Traditional search ranked whole pages against queries — a page either matched well enough to rank, or it didn't, judged as a single unit. Passage indexing, which Google introduced in 2021, changed that: individual passages within a page can now be surfaced for specific queries even when the page's overall focus is something else entirely.

What this changes for AEO

For AEO, passage indexing means a page's value isn't determined solely by its overall topic match anymore — any well-structured section inside it can be extracted and cited independently:

• A long-form blog post with a well-structured FAQ section can have that FAQ cited in an AI Overview even if the blog post as a whole doesn't rank for the specific FAQ query
• An About page with a specific founding year and credential mention can have that exact passage extracted for credential-specific queries, without the rest of the page being relevant at all
• A service page with a clearly structured "how it works" section can have that passage extracted for process-related queries, independent of whatever else the page is selling

Why this matters more for AI systems than it did for classic search

Classic passage indexing was already useful for search, but AI systems push the logic further. A model answering a user's question isn't reading a whole page and summarising it — it's retrieving the specific passage most likely to answer that exact question and building its response from that, often stitching together passages from several different sources. A page where every idea depends on the paragraph before it is much harder for that retrieval process to use well than a page built from self-contained sections.

Practical implications for content structure

Because AI systems and passage indexing both extract at the passage level, every section of a page should be independently readable and informative — not dependent on the surrounding sections for context. A section that opens with "as mentioned above" or "building on this" is harder for a retrieval system to lift cleanly than one that restates just enough context to stand alone.

This reinforces the answer capsule → approach directly: make every section's first paragraph a self-contained, directly extractable answer to whatever implicit question that section is addressing, rather than saving the actual answer for the third sentence after some throat-clearing.

Common mistakes

The most common mistake is writing sections that build narratively on each other — useful for a human reading top to bottom, useless for a system trying to extract just the third section in isolation. The second is burying the actual answer to an implied question a paragraph or two into a section, after a sentence or two of setup; a passage-extraction system tends to grab the opening of a section, and if that opening is warm-up rather than substance, the extracted passage is the warm-up.

Adjacent concepts

FAQPage schema → and Speakable schema → both work in the same direction as passage indexing — marking specific chunks of a page as independently extractable, structurally rather than just through good writing. Query fan-out → describes how a single user question can expand into several retrieval passes across a page or a site, each one potentially pulling a different passage — which is part of why every section, not just the obviously important ones, benefits from being self-contained.

Related terms: Answer Capsule → · FAQPage Schema → · Speakable Schema → · AEO → · Query Fan-Out → · Structured Answer →

Example: A Hyderabad-based orthodontics clinic's blog post on "how much do braces cost in India" was written mostly as a general overview, but included one tightly written paragraph directly answering "how long does Invisalign treatment usually take." That single paragraph, not the post as a whole, started appearing in AI Overview answers to Invisalign-duration queries — a topic the post wasn't primarily about, extracted because that one passage happened to answer the question cleanly on its own.