diff --git a/docs/CLAIM-REVIEW-CYCLE.md b/docs/CLAIM-REVIEW-CYCLE.md
new file mode 100644
index 0000000..d9e50e5
--- /dev/null
+++ b/docs/CLAIM-REVIEW-CYCLE.md
@@ -0,0 +1,51 @@
+# Claim Review Cycle
+
+Open Problem Lab should not treat accepted claims as permanently current.
+
+This guide defines a simple review cycle for claims whose usefulness can decay over time because the evidence base, field setting, policy environment, dataset, or implementation context changes.
+
+## Why this matters
+
+A claim can be correct at merge time and misleading later.
+
+The repository needs a visible way to say when a claim should be rechecked, refreshed, bounded, or archived. This is especially important for claims used in prioritization or field-facing decisions.
+
+## Minimum review-cycle note
+
+```markdown
+## Claim review cycle
+
+**Claim:** Link to claim, evidence record, issue, or PR.
+
+**Current status:** accepted | bounded | disputed | field-pending | replication-pending | archived.
+
+**Review cadence:** on challenge | 6 months | 12 months | 24 months | event-triggered.
+
+**Why this cadence:** One sentence.
+
+**Refresh trigger:** Source update, dataset release, policy change, field report, replication result, or reviewer concern.
+
+**Owner:** maintainer | reviewer | agent | generated index | external source.
+
+**Action if overdue:** recheck | mark bounded | mark field-pending | archive | open issue.
+```
+
+## Cadence guidance
+
+- Use `on challenge` for stable background claims.
+- Use `24 months` for slow-changing structural facts.
+- Use `12 months` for program status, burden estimates, or institutional context.
+- Use `6 months` for claims tied to fast-moving operating environments.
+- Use `event-triggered` when a known update source controls freshness.
+
+## Reviewer guidance
+
+Reviewers should ask for a review cycle when a claim is likely to become stale but is being used as durable support.
+
+## Agent guidance
+
+Agents should not reuse an accepted claim without checking whether its review cycle is overdue.
+
+## Acceptance test
+
+A review cycle is adequate when a future contributor can answer when the claim should be checked again and what to do if it is overdue.