Chapter 33: Claims Integrity & Content Governance

Last Updated: 2026-04-12

33.2 When Claims Integrity Applies

Claims Integrity review is required before:

1. Publishing any marketing content (blog posts, social media, campaigns)
2. Publishing portfolio documents
3. Releasing product descriptions or feature lists
4. Quarterly review of all existing marketing materials

Note: Claims integrity applies to any factual assertion — not just explicit statistics. “Our products save time” is a claim that needs evidence just as much as “increases conversions by 40%.”

---

33.3 Running a Claims Integrity Audit

“Run a claims integrity audit on [document name or content].”

Or apply the skill directly:

“Apply the claims-integrity-audit skill to [document name or content].”

The Claims Ombudsman agent (or the claims-integrity-audit skill):

1. Identifies every factual claim in the document.
2. Categorizes each claim (performance claim, expertise claim, outcome claim, feature claim).
3. For each claim, asks: “What evidence supports this?”
4. Flags claims that are:

• Unsupported — no evidence exists
• Overstated — evidence exists but doesn’t support the strength of the claim
• Stale — evidence is outdated (data more than 12 months old)
• Legally risky — superlatives or guarantees without qualification

1. Provides a remediation recommendation for each flagged claim.

---

33.4 The Claims Evidence Registry

All claims used in marketing are tracked in a Claims Evidence Registry. This is a structured document maintained by the Claims Evidence Curator agent or built using the claims-evidence-registry skill.

Registry format

# Claims Evidence Registry — [Document or Campaign Name]

## Claim Registry

### CL-001
**Claim:** "Our AI workflows reduce content production time by 60%."
**Type:** Performance claim
**Evidence:**
  - Source: Internal time study, March 2026
  - Methodology: Compared pre-AI and post-AI content production time for 3 clients over 90 days
  - Result: Average 58% reduction (rounded to 60% in marketing copy)
**Confidence:** High
**Staleness Threshold:** 12 months (review by March 2027)
**Status:** Active

### CL-002
**Claim:** "Built across multiple domains with AI working fluency."
**Type:** Capability / evidence-base claim
**Evidence:**
  - Source: Active builder portfolio (see Chapter 32)
  - Note: Aggregate claim — tied to the shipped project inventory, not client names
**Confidence:** High
**Staleness Threshold:** 24 months
**Status:** Active

---

33.5 Staleness Checks

Outdated evidence weakens claims and creates legal exposure. The Claims Evidence Curator runs quarterly staleness checks.

“Run a staleness check on claims evidence.”

The curator reviews every registry entry and flags claims where:

• The evidence is older than the staleness threshold
• The underlying reality may have changed (market stats, product capabilities)
• New contradicting evidence has emerged

For each stale claim, the options are:

1. Update the evidence — gather fresh data to support the claim.
2. Revise the claim — weaken or remove the claim if fresh evidence is unavailable.
3. Remove the claim — delete from marketing materials.

---

33.6 Content Publication Gate

The publication gate is a checklist run before any content goes live:

□ Claims Integrity audit completed
□ All claims have evidence in the registry
□ No stale evidence (all evidence within staleness threshold)
□ No legally risky language (unqualified superlatives, guarantees)
□ Fact-checking skill applied to any factual assertions about the market or industry
□ Brand voice is consistent with brand guidelines
□ Approved by Peter Westerman (or designated reviewer)

---

33.7 AI-Generated Content Policy

All ITI marketing content, including AI-assisted content, must pass the same claims integrity review. AI-generated text is particularly prone to:

• Hallucinated statistics — plausible-sounding numbers that aren’t real.
• Overstated capabilities — describing features or integrations that don’t exist.
• Outdated information — Claude’s training data has a cutoff date.

Rule: Never publish AI-generated factual claims without independent verification. Always apply the fact-checking skill before publishing AI-drafted content. For claims about capabilities or outcomes, also apply the claims-integrity-audit skill.

---

Previous: Chapter 32 — Product Portfolio | Next: Appendix A — Glossary