How it works
Paste any content and the checker computes a composite verifiability score from 8 metrics: external citations, source attributions, numeric facts, dates, named entities, direct quotes, hedge density and opinion framing. The last two are inverted - lower is better.
Why these eight metrics
- External URLs give assistants a direct path to verify - without them, every claim is a dead end.
- Attributions ("according to", "per", "source:") signal transparency - assistants reward transparent content.
- Direct quotes with a named speaker are the highest-signal evidence class - they're lifted verbatim 6x more often than paraphrased claims.
Where to fix
For any metric under 50, work through the list: add a URL, add a date, add a named source. Combine with the Claim Verification Checker for sentence-level rewrites, and the Trust Signal Analyzer for the broader signals pass.