Documentation
Limitations
This tool is useful. It is also limited. Knowing where it stops being useful is part of using it well.
What this tool DOES
- Surfaces fact-check reviews published by major IFCN-verified outlets, when they exist.
- Indexes news coverage from the last ~30 days across 65+ languages via GDELT.
- Looks up encyclopedia background via Wikipedia.
- Scores the editorial credibility of the outlets covering a topic.
- Detects spoofed domains, obvious clickbait, and missing bylines on URLs you submit.
- Flags claim types that need expert review (medical, legal, financial).
- Shows you exactly which APIs returned which evidence — every card has a real, clickable URL.
What this tool DOES NOT do
- It does not establish truth. The verdict reflects what fact-checkers said — not what is true in some absolute sense.
- It cannot read paywalled or login-gated articles.
- It cannot verify future predictions.
- It cannot evaluate opinions, value judgments, or contested definitions.
- It cannot detect novel disinformation that has not been fact-checked anywhere.
- It cannot replace a doctor, lawyer, financial advisor, or any other professional.
- It cannot guarantee that an outlet is unbiased — only that its track record is documented.
Known data gaps
- GDELT only indexes the last 30 days. Older claims will return no news evidence.
- Google Fact Check Tools coverage is uneven. Specific phrasings matter; rewording the claim can change which fact-checks match.
- Local source rules cover ~80 outlets. Anything outside that list gets TLD-only inference.
- Stance reconciliation is heuristic. Fact-check ratings that don't materially match your claim are demoted to “related,” not used as a verdict.
- Big publishers often block bots. If we can't fetch a URL you submitted, domain-level signals are still shown.
- English-language only. Non-English claims work only if matching English coverage exists.
What we will NEVER do
- Use mock data, fake demo results, or fake trending topics.
- Iframe arbitrary external sites — most block embedding, and it can hide source identity.
- Use AI to invent quotes, sources, or stances. The verdict comes from fact-checker output, not language-model guessing.
- Require an account, payment, or share your queries with third parties beyond the public APIs we consult.
- Claim a result is “100% true” or “100% false” or use “AI truth detector” framing.
When you should NOT rely on this alone
- Medical decisions — talk to a licensed clinician.
- Legal decisions — talk to a licensed attorney.
- Financial decisions — talk to a licensed advisor.
- Time-sensitive breaking news that may still be developing.
- Topics where you need the primary source (a study, ruling, or filing) — find that, not a news summary.
Use this tool as one input in your verification process, alongside primary sources and your own judgement. It is a transparency layer for source quality, not an oracle.