Advice is cheap.
Evidence is the price.
Paste any claim, piece of advice, or AI answer. ClaimLens splits it into checkable parts, scores every part on evidence signals you can inspect, and tells you what would change its mind.
Verdict
Score = transparent rubric, not a model. Every point is explained below — no black box.
Evidence map — each claim as a point in 3D
Height = confidence score. Taller, brighter points are better-evidenced. Select a point (or a row below) to inspect it. Drag to tilt the map.
Claim by claim
Decision brief
Past analyses
How the score is built — nothing hidden
| Signal | Points | Why it counts |
|---|---|---|
| Specific number or statistic | +15 | Checkable quantities can be verified or falsified. |
| Date or year given | +10 | Anchors the claim in time; stale claims can be caught. |
| Named source / citation | +10 | “According to X” gives you somewhere to check. |
| Calibrated hedging (“may”, “likely”, “suggests”) | +10 | Honest uncertainty is a sign the speaker knows the limits. |
| Absolute words (“always”, “never”, “proves”, “guaranteed”) | −15 | Reality rarely deals in absolutes; overclaiming is a red flag. |
| Weasel attribution (“experts say”, “studies show”) | −10 | Authority with no name cannot be checked. |
| Hype words (“miracle”, “revolutionary”, “best ever”) | −10 | Marketing language crowds out evidence. |
| Too short to check (< 6 words) | −5 | A bare assertion carries no evidence at all. |
| Number/statistic with no source or date to check | −10 | A precise figure with nowhere to check it is a claim, not evidence. |
Base score 50, clamped 0–100. Verdicts: ≥ 70 Solid · 40–69 Shaky · < 40 Unverifiable. ClaimLens is a screening tool, not a fact-checker: it measures how checkable and how overclaimed a statement is. It makes no web requests and calls no model — everything runs in your browser, and nothing you paste leaves your device.