Lessons 72.2, 73.4, and 79.2 each demanded a preflight — and each watched fluent creative quietly overstate the offer. This appendix is the printable checklist: message match, proof, call to action, accessibility, and tracking, ending in a pass, fix, or hold verdict per asset.

The checklist (run per asset + destination pair)

Message match. Does the headline, promise, price, date, and availability on the creative continue verbatim on the landing page? Does the page serve the exact audience the creative addressed?

Proof. Is every claim beside evidence — screenshot, source link, real testimonial with permission, current numbers with dates? Are AI-generated visuals, voices, or quotes labeled where the audience would care?

Call to action (CTA). One action, visible without scrolling on mobile, repeated after the proof. Does the button say what happens next ("Save the template," "Book Friday slot") rather than "Submit"?

Accessibility (a11y). Readable contrast, captions checked line by line, no essential meaning carried by color or flashing alone, alt text that teaches, keyboard-reachable action. If the asset excludes someone, it fails regardless of beauty.

Tracking. One destination URL with readable source labels, one conversion event defined as first value (not click), owner named, review date set. No tracking that the privacy promise did not disclose.

Printable form

# AD-AUDIT.md — [Campaign], date: ___ / auditor: ___

## Asset 1 — [name + link]
- [ ] Message match (promise/price/date/audience continues): ___
- [ ] Proof beside every claim (dated): ___
- [ ] CTA: one action, wording: ___ / mobile visible: ___
- [ ] A11y: contrast / captions / alt / no color-only meaning: ___
- [ ] Tracking: URL labels / event = first value / owner / review date: ___
- Verdict: PASS / FIX / HOLD — reason: ___ / owner: ___ / fix by: ___

## Asset 2 — (same) / Asset 3 — (same)

## Destination — [page URL]
- [ ] Same five checks on the page itself: ___
- Verdict: PASS / FIX / HOLD: ___

Verdicts, honestly used

• Pass: ship. Minor polish may continue after launch, logged, not blocking.
• Fix: ship after named changes by a date — new screenshot, corrected price, captioned video, labeled tracking. List the fixes; re-check only the fixed lines.
• Hold: do not ship. Misleading promise, broken destination, missing permission for a testimonial, inaccessible core message, or untracked spend. State what must change before a second review.

Use the two-person pattern where possible: maker checks craft, a second person checks claim and customer interpretation. The second reviewer asks one question throughout: "Would a stranger who believes exactly what this says be rightly served on arrival?"

Worked mini-example

Research Desk ad plus template page: message match fix — creative says "free filing check" but page leads with pricing; move the free template above the fold. Proof pass — sample brief with dated sources beside the claim. CTA fix — button reads "Submit"; rewrite to "Run the check on my company." Accessibility hold on the video variant — auto-captions unchecked and key step conveyed by color only; correct captions and add labels before spend. Tracking pass — source labels plus "first brief saved" event, owner and Friday review set.

Verify fast

Pick any verdict. Can you point to the exact line, screenshot, or timestamp that earned it? A verdict without evidence is an opinion. Opinions do not survive contact with spend.