Source Classification — Primary and Secondary Evidence

Every evidence record on this site carries an authority field with exactly two possible values: primary or secondary. This page documents what each means and how it affects whether an entity or comparison can be publicly indexed.

Primary

A source is primary when it originates from the entity's own official channel: its GitHub/GitLab repository (via the official API), its own model card, its own documentation, or its own official site. Examples already in this catalog: api.github.com/repos/... for repository identity, huggingface.co/api/models/... for model cards, a vendor's own pricing or feature page.

Only primary-authority evidence with a verified status and a validly-formatted content hash counts toward the publication gate (isEntityIndexable). A primary source can still turn out to be wrong or outdated — primary describes where the claim came from, not a guarantee of permanent accuracy.

Secondary

A source is secondary when it reports on or analyzes an entity rather than originating from the entity itself — independent reporting, third-party benchmarks, or community-maintained documentation. Secondary evidence is recorded in the same EvidenceRecord shape but does not, on its own, satisfy the publication gate for verified status.

There is no third tier

Some earlier drafts on this site described a third community tier (forum posts, social media). That tier was never implemented — the real EvidenceAuthority type in lib/catalog-types.ts only defines "primary" | "secondary". Anecdotal sources aren't a recognized evidence class on this site at all; they're simply not evidence.

Related methodology pages

Source receipts

Internal — lib/catalog-types.ts (EvidenceAuthority, EvidenceRecord). The real type definitions this page describes.
Internal — lib/catalog.ts (isEvidenceVerified). The function that requires authority === "primary" for verified status.

Last reviewed

Page last reviewed: 2026-09-02