Cite authoritative sources inline
What the scan checks: it counts outbound links inside your main content (navigation and footer links are excluded, because a partner logo in the footer is not a citation) and requires:
- at least two external sources, and
- at least one of them from an authoritative tier.
The tiers we recognise today:
| Tier | Examples |
|---|---|
| Government or intergovernmental | any .gov, .gov.uk, .mil, europa.eu, WHO, UN, OECD, World Bank |
| Academic | any .edu, .ac.uk and the other ac.<cc> domains |
| Reference | Wikipedia, Wikidata, Britannica |
| Standards body | W3C, IETF, RFC Editor, ISO, IEEE, IANA, WHATWG, Ecma, Unicode |
| Peer-reviewed or research | Nature, Science, ScienceDirect, Springer, arXiv, JSTOR, DOI, The Lancet, BMJ, NEJM, ACM, PNAS, Cochrane |
| Publication of record | Reuters, AP, BBC, NYT, WSJ, FT, The Economist, Bloomberg, Washington Post, The Guardian, NPR, HBR |
Fewer than two external sources is a fail. Two or more with none from those tiers is a warning, and the report names the hosts you did cite. On YMYL topics (money, health, law, safety, civics) the authoritative source is required rather than preferred, so the same page is a fail there.
The list is deliberately short and is a starting point. If a source you consider authoritative in your vertical is missing from it, that is a gap in our list, not a verdict on your page: tell us and it gets added.
Why it matters for AI search visibility
Section titled “Why it matters for AI search visibility”An answer engine is assembling a claim it will be held to. Content that cites primary sources gives the model something to corroborate against, and models weight corroborated claims more heavily than uncorroborated ones. The manual’s criterion states it plainly: at least two outbound links to .gov, .edu, Wikipedia or genuine industry sources.
The reason the tiering matters is that the old version of this check counted any two external hosts, so a page whose only outbound links were to two other marketing blogs scored the same as a page citing the CDC and a peer-reviewed study. Those are not the same page, and an engine does not treat them as the same page.
The fix
Section titled “The fix”- Cite where the claim is, not in a “Sources” dump at the bottom. Inline citation ties the source to the specific sentence it supports, which is what survives when the chunk is lifted.
- Go to the primary source, not the write-up of it. If your number comes from a study, link the study. If it comes from a regulation, link the regulation. Linking a blog post that links the study puts a stranger between you and the evidence.
- On regulated topics, make it dense. The manual’s YMYL rule: every legal claim cites the statute, regulation or case; every medical claim cites a peer-reviewed source or a government health authority; every numerical claim cites the underlying data.
- Do not fake it with a Wikipedia link. One reference-tier link plus a body of unsourced assertions passes this check and fails the reader. The check is a floor, not a target.
What this check does not do
Section titled “What this check does not do”It does not judge whether the source actually supports your claim, and it does not fetch the links to see whether they still resolve. Both are real quality questions; neither is answerable from the page’s own markup, which is all a scan reads.