Source hierarchy
Articles prefer, in rough order: the Oxford English Dictionary; Merriam-Webster; Cambridge Dictionary; Oxford Learner’s Dictionaries; historical dictionaries; Google Books, Internet Archive, and HathiTrust scans; newspaper archives; library and university collections; digitized primary texts; peer-reviewed linguistic or historical research; and established etymology references.
General-interest blogs, anonymous idiom sites, discussion boards, and commercial marketing pages may document a popular story or current usage, but they are never the only basis for a strong historical or medical claim.
How claims are checked
Claims that can be checked against a dated printed source are checked that way. When a citation cannot be found, the article keeps the useful topic, rewrites categorical language into an honest uncertainty statement, and labels the story as folklore or an unverified theory.
Source-quality review
Every external source domain used by the library is classified on a three-tier scale: reference/authoritative (dictionaries, archives, primary texts, established reference works), popular (needs corroboration), or weak/user-generated (never a sole basis for a historical or medical claim). The classification is applied by the publisher during research and recorded in the editorial workflow; it is an internal working tool, not a public file.
In practice this means an article may cite a popular idiom site — but only to document that a story circulates, never as the sole support for a strong claim. Where the evidence is thin, the article weakens the claim and labels the uncertainty instead of repeating the story as fact.
What to do if a source is wrong
Report it through the corrections page. The article is rechecked and the correction is logged.