Local claim-check review workspace

Check AI answers for unsupported claims and overconfident wording.

Paste an answer, extract likely claims, mark what is supported, and use the rewrite panel to produce safer language before publishing.

Disclaimer: This is an assistive review tool. It flags risk signals and reviewer decisions, not definitive truth.
Evidence-first review desk

Move every answer through a clean evidence pass before it ships.

Separate claim extraction, source confidence, and safer rewrite decisions so reviewers can move quickly without overstating certainty.

  • Source-backed
  • Rewrite guardrails
  • Review history

AI answer input

Claims are extracted with deterministic text heuristics in this browser.

Status defaults to “needs source” unless a source marker is present.

Honesty score

Deterministic risk score

0
Based on unsupported claims, needs-source claims, source markers, hedging, and absolute wording.
Claims
0
Unsupported
0
Supported
0
Needs source
0
Score math
Status penalty
-0
Absolute penalty
-0
Source marker bonus
+0
Hedging bonus
+0

Claim table

Edit each status as you review sources outside this tool.

Paste an AI answer to begin.

Selected claim guidance

Select a claim after extraction to see reviewer guidance.

Recommended rewrite panel

Use this checklist to rewrite unsupported or unsourced assertions before publishing.

Paste an AI answer and run a review to generate a rewrite checklist.

Feedback routing

Use the floating feedback widget to turn unclear claim states, missing evidence categories, or rewrite problems into GitHub issues with page diagnostics.

GitHub target: alibad/honestai