The Vault aims to be a genuinely useful reference, not just a repository of claims — which means the quality of what backs up a statement matters as much as having a citation at all. This page sets out how to judge whether a source is good enough to cite, and how to handle it when sources disagree.
The short version
- Prefer sources that are independently checkable, specific, and attributable to a named author, publication or institution over anonymous or unattributed ones.
- Corroborate specific factual claims — names, dates, figures, quotes — across more than one independent source where possible, especially for anything surprising or load-bearing in the article.
- Be explicit in the article when sources genuinely disagree, rather than silently picking one and hiding the discrepancy.
- Treat AI-generated reference sites as a starting point for leads, never as a citation in their own right.
- A source existing and being reachable is not the same as it being reliable — check what it actually says, not just that the link resolves.
Source quality, roughly ranked
Strongest:
- Primary sources: official statements, court records, original research papers, an organisation's own published material about itself (used carefully — organisations aren't neutral about themselves, but they're authoritative on their own basic facts like founding dates).
- Established news organisations with a track record of corrections and editorial standards (BBC, major national newspapers, wire services).
- Specialist reference works with named editorial oversight — the Psi Encyclopedia (Society for Psychical Research), Encyclopædia Britannica, and similar.
- Peer-reviewed academic sources, where available.
Generally fine, used with normal care:
- Wikipedia — genuinely useful as a well-corroborated starting point and for its own citation trail, but summarise in your own words per the Copyrights policy, and where a claim really matters, follow Wikipedia's own citations back to the original source rather than citing Wikipedia as the final word.
- Retailer and database listings (Amazon, IMDb, Goodreads, publisher pages) — fine for factual, checkable details like release dates, ISBNs, cast lists and page counts, which these are well-suited to getting right. Less reliable for anything requiring editorial judgement, like critical assessment.
- Established fan sites and archives with a track record of accuracy and citation of their own (e.g. Higgypop for Most Haunted episode data) — useful for niche detail no mainstream source bothers to record, but corroborate anything that seems significant.
- Individual reviews (Goodreads, blogs, contemporary press reviews) — fine for representing "a reviewer said X," properly attributed, but shouldn't be presented as if they reflect a consensus unless several independent ones agree.
Use with real caution:
- AI-generated reference sites (Grokipedia and similar) — these can be a useful lead for facts to go and verify elsewhere, but should not be relied on as the sole citation for a claim, and never for anything contested or surprising. If a fact only turns up on an AI-generated source and nowhere else, treat that as reason for more searching, not less.
- Personal or hobbyist sites with no clear editorial standard, especially where a claim can't be corroborated anywhere else — proceed, but flag internally (or in the article, if the claim is doing real work) that it rests on a single, less-established source.
- Sites making check-worthy claims where you genuinely can't verify their content directly (e.g. blocked by robots.txt, requiring login) — note the limitation honestly rather than assuming the claim holds up.
Don't cite at all:
- Anonymous forum posts or social media claims with no way to verify the poster's basis for the claim.
- A source's own unsupported claim about itself where independent verification exists and contradicts it.
When sources disagree
Sources will sometimes genuinely conflict — a disc count that's "3" in most listings and "4" in one, a birthplace given differently across two obituaries, a quote rendered slightly differently in two outlets. When this happens:
- Don't silently pick whichever source you found first, or whichever supports the more interesting version of the story.
- Where possible, favour the reading that's corroborated by multiple independent sources over a single outlier.
- Where it's a genuinely open question, say so in the article rather than presenting false certainty — a short note like "sources differ on X; most give Y" is honest and useful to a reader, especially where a discrepancy might reflect a real underlying ambiguity rather than an error.
Checking a source properly
Before citing something, especially for a specific fact rather than general framing:
- Read what the source actually says, not just its title or a search snippet — a source can exist and be genuine while still not actually supporting the specific claim being attached to it.
- Be skeptical of your own first search result. A source not being findable doesn't necessarily mean it doesn't exist (and vice versa, a source existing doesn't automatically mean its specific content backs up the claim) — where it matters, try to actually read the page rather than infer from a listing or blocked-access error.
- Where a claim is genuinely important to an article's credibility (a controversy, an exposure, a contested figure), it's worth the extra search or two to corroborate it independently rather than resting on one source alone.