The Source File: Why AI Models Cite Some Pages and Skip Others

What "source file" means in AEO

When an AI model answers a query, it does not cite every page it has access to. It cites a small subset. The selected subset is the source file for that query. In OnlyAEO's data, the source file is typically three to seven pages per query, even when the model has indexed dozens of candidate pages.

Getting into the source file is the goal of AEO. Once a page is inside it, citations on the query are durable. The page becomes a default citation source. The brand's compounding accelerates because every subsequent piece on adjacent topics inherits some of the source-file equity.

The five attributes of source-file pages

Across hundreds of citation audits, the source-file pages share five attributes consistently. The combination matters more than any single one.

1. A clear answer surface. An answer capsule, a single-line direct answer, or a structured definition at the top of the page. The page tells AI exactly what claim it can extract.

2. Structured data. FAQPage schema, Article schema, or HowTo schema. Pages without schema are still cited occasionally, but at lower rates than otherwise-equivalent pages with schema.

3. Named authorship. A named human author with a credential line and at least one public sameAs profile. Anonymous authorship is the largest single source of source-file exclusion in OnlyAEO's data.

4. Consistent entity signal. The brand publishing the page has a coherent entity profile: Crunchbase, LinkedIn, sameAs network, and consistent claims about itself across all of them.

5. Topical authority on the specific query. The page is part of a cluster of related pages on the same topic, with internal linking that signals the cluster to AI. A standalone page on a topic the brand has not otherwise covered rarely makes the source file.

A page with all five attributes is in the source file for its query, almost always. A page with three of five sometimes makes it. A page with two or fewer rarely does.

The source-file selection process

The selection process AI runs (as best as observable from outside the model) follows roughly this sequence.

The filtering stage is where most pages drop out. A brand with weak structural signals never reaches the ranking or extraction stages, regardless of how strong the underlying content is. Fixing structural signals is therefore the most leverage-rich AEO investment.

How brands move into the source file

The OnlyAEO playbook for moving a brand into the source file for a target query follows a fixed sequence.

Step one is the answer surface. Open the target page with a clear answer capsule. Step two is structured data. Add FAQPage and Article schema. Step three is named authorship. Rebrand the author from a team byline to a named human with a credentialed profile. Step four is cluster building. Publish three to five related pieces and link them inbound to the target page.

The four steps together usually move a page from outside the source file to inside it within six to eight weeks. The mechanism is structural, not magical. The page now meets the criteria the filtering stage screens on.

Why being inside the source file matters more than topic quality

A surprising finding from OnlyAEO's audits: a thinner page inside the source file usually earns more citations than a deeper page outside it. The reason is the filtering stage. Deep content that fails on structural signals never gets to compete on depth. Thinner content that passes structural signals competes and often wins.

The implication is that brands optimizing for content depth alone, without addressing structural signals, are usually leaving citation lift on the table. The right investment order is structural signals first, then content depth. The reverse order produces fewer citations per editorial hour.