TL;DR
- This is a publish-ready verification checklist for AI-assisted drafts.
- It covers claims, evidence, sources, and reasoning.
- Use the acceptance rule: publish, escalate, or rewrite.
Not AI search monitoring. Not authorship detection. This framework verifies what’s true and supported inside your draft: facts, claims, sources, and reasoning.
Layer 1 — Claim extraction
Goal: Identify what the draft is asserting (facts, stats, comparisons, recommendations, causal statements).
Failure modes: vague claims, hidden assumptions, overconfident language (“always,” “proven,” “best”).
Go deeper: Claim verification →
Layer 2 — Evidence validation
Goal: Determine whether high-impact claims are supported by evidence.
- Stats: verify source + timeframe + method.
- Comparisons: confirm units and baselines match.
- Causal claims: require stronger evidence than correlation.
Layer 3 — Source authentication & alignment
Goal: Ensure sources exist, are credible, and support the exact claim.
Failure modes: fake citations, misquoted sources, outdated references, source/claim mismatch.
Go deeper: Citation verification →
Layer 4 — Reasoning & consistency checks
Goal: Catch contradictions, logic leaps, missing caveats, and definition drift.
Go deeper: Reasoning checker →
Layer 5 — Risk calibration & acceptance rule
Use this acceptance rule:
- ✅ Publish/Share if high-impact claims are validated and uncertainty is labeled.
- ⚠️ Escalate if a high-impact claim can’t be validated quickly.
- 🛑 Rewrite/Remove if citations are fake/misaligned or conclusions don’t follow.
Copy/paste checklist
- [ ] Extract claims (facts, numbers, comparisons, recommendations)
- [ ] Verify Tier-A (decision-changing) claims first
- [ ] Confirm sources exist and support each claim
- [ ] Check reasoning: contradictions, leaps, missing caveats
- [ ] Apply acceptance rule (publish / escalate / rewrite)
Use it by document type
Want to see how Clarity operationalizes this? How Clarity works.