How it works
Paste any content and the tool extracts every declarative sentence, then classifies it as verifiable, partially verifiable or unverifiable. Verifiable = assistants can check it against a known source; partially = contains some facts but lacks attribution; unverifiable = hedged or opinion-form.
The three failure modes
- Hedged claims ("might", "may", "probably") - assistants treat these as low-confidence and skip them.
- Opinion framing ("I think", "in my opinion", "arguably") - gets extracted as sentiment, not fact.
- Missing attribution - a strong claim without a source gets downgraded even when it's correct.
Fix pattern
For every unverifiable claim: add a source, remove the hedge, replace opinion framing with a declarative verb. Pair with the Evidence Strength Analyzer and the Fact Density Analyzer. Full playbook: write content AI assistants cite.