How it works
Proofbase is a source-quality scanner. It does not decide truth — it consults public APIs and a local rules database, then shows you what each source said and how credible those sources are.
Data sources
Indexes fact-check articles published by Politifact, Snopes, FactCheck.org, AFP, Reuters Fact Check, and dozens of other IFCN-verified outlets. When matching fact-checks exist for your claim, their textual rating drives the verdict label.
Limit: Free quota covers thousands of queries per day. Without an API key, this source is silently skipped.
Global news index covering tens of thousands of outlets in 65+ languages. We use it to find recent news coverage of a claim or domain. GDELT items are labeled 'Related' — we never infer a verdict from coverage alone.
Limit: GDELT only covers the last 30 days of news.
Wikipedia's MediaWiki Search + REST summary endpoints. Used for background context — Wikipedia entries never drive the verdict.
Limit: Community-edited — orientation tool, not a primary source.
A hand-curated JSON file (data/source-rules.json) listing ~80 publishers with their category, base quality score, warning flags, and recommended use. Guides quality scoring — does not decide truth automatically.
Limit: Unlisted domains fall back to TLD heuristics (.gov / .edu trusted, .info / .xyz penalized).
What the verdict labels mean
- Evidence suggests this is supported — Multiple reputable sources report findings consistent with the claim. Not a guarantee of truth.
- Evidence suggests this is disputed — Fact-checkers on balance rated the matching claim as false, misleading, or unsupported. Not a guarantee of falsehood.
- Conflicting reports detected — Reviewers reached different conclusions. The claim likely depends on framing.
- Limited evidence found — Either no fact-checker has rated this phrasing, or no source returned anything. Absence is not a verdict.
What the Source Quality Score means
The 0–100 score is a measure of outlet credibility — the average editorial track record of the publishers covering your topic, with bonuses for fact-checker presence, source diversity, and recency.
It does notmeasure how likely the claim is to be true. A claim covered by Reuters and the New York Times gets a high score whether the claim itself is right or wrong — what we're saying is “these are reputable outlets discussing this,” not “this is true.”
What the tool cannot prove
- It cannot tell you a claim is “100% true” or “100% false.”
- It cannot read paywalled or login-gated articles.
- It cannot evaluate predictions about future events.
- It cannot replace a doctor, lawyer, or financial advisor.
- It cannot detect novel disinformation that no fact-checker has covered.
- It cannot tell whether sources are biased in ways their published track record hides.
Ready to try?
Paste a claim, a URL, or a domain on the home page.
Start a check →