Audit marketing copy for exaggeration and evidence gaps
Trace factual and implied claims to evidence before copy is published
9 min setupTested with:ChatGPTReviewed: 2026-08-28
1
Add your context
Your text stays in this browser. AILesson Prompts does not send it to a model or server.
2
Your prompt
Unfilled fields remain visible as placeholders, so you can still copy and edit the prompt
Audit the exact marketing copy against only the supplied evidence and rules.
Copy, channel, audience, market, placement, CTA, and publication date:
[copy]
Product facts, studies, samples, dates, methods, approvals, limits, and locators:
[evidence]
Legal or brand rules, prohibited terms, qualifiers, comparison basis, and tolerance:
[rules]
Extract every explicit and reasonably implied factual, quantitative, comparative, superlative, testimonial, performance, outcome, urgency, availability, and guarantee claim. Preserve an exact excerpt and locator. For each, record likely audience interpretation, evidence source and match of population, measure, conditions, timeframe and date, status as supported, partially supported, unsupported, contradictory, or unverifiable, and risk with rationale. Detect missing baselines, denominators, sample context, qualifiers, material limitations, and ambiguous words. Do not decide legal compliance; flag items for qualified review. For risky items recommend the smallest correction: substantiate, qualify, narrow, attribute, remove, or convert to a testable opinion. Never create evidence or a stronger replacement claim. End with release blockers, clarification questions, and a publish, revise, or hold recommendation.
Private by defaultPrompt assembly happens locally in your browser. Avoid placing confidential information into any AI service unless your organization allows it.
From input to outcome
A worked example
See how concrete context turns this recipe into a usable result
Actual input
Marketing copy and context
Hero: The fastest way for every team to close the books. Cut month-end work by 50%—guaranteed. Trusted by finance leaders everywhere. Start today; offer ends Friday. CTA: Save half your time.
Evidence package
Product automates reconciliation suggestions. One customer study: 12-person US software-company finance team reduced median reconciliation task time from 20 to 13 hours across three month-end cycles, measured internally in May-July 2026. Customer approved: 'The review queue made our handoffs easier to see.' Product available in US and Canada. No guarantee program, market-share survey, global usage evidence, or approved promotion end date.
Claim rules
US homepage, publish Sep 1, 2026. Brand prohibits 'guaranteed' without written program terms and 'fastest' without comparative testing. Numerical claims must identify result context. Testimonials must be exact and approved. Legal reviews performance and urgency claims.
Example output
Recommendation: hold and revise before legal review.
Claim ledger
“fastest” — comparative superlative, unsupported and prohibited without comparative testing. High-risk release blocker. Remove; no supplied evidence establishes rank against alternatives.
“every team” — universal applicability claim, contradicted by evidence limited to one 12-person US software-company finance team and availability in two countries. High risk. Narrow to a supported product capability and eligible market; do not generalize the case.
“Cut month-end work by 50%” and CTA “Save half your time” — quantitative performance claims, unsupported as written. The study concerns reconciliation task time, not all month-end work, and the observed reduction from 20 to 13 hours is 35%, not 50%. High-risk blocker. If approved, attribute and qualify the exact 35% result, task, company context, three-cycle period, internal measurement, and study dates.
“guaranteed” — explicit guarantee with no program. Unsupported and prohibited. Remove; legal review required.
“Trusted by finance leaders everywhere” — broad popularity, geography, and audience claim. No survey or global evidence. High risk; remove. The approved quote supports only easier-to-see handoffs for one customer.
“offer ends Friday” — urgency and deadline claim, unverifiable because no promotion date or terms were supplied. High-risk blocker. Confirm exact date, timezone, eligibility, and terms or remove.
Implied CTA promise — clicking will save each visitor half their time. Unsupported for the same population and measure reasons as item 3.
Clarify whether the customer study and attribution wording are approved for public use, whether “reconciliation suggestions” requires human review disclosure, and whether any promotion exists. A revised page may describe the verified capability and exact approved testimonial now; performance copy should remain on hold until legal approves a properly scoped study statement.
Why this works
1
Auditing implied claims catches promises created by context even when no sentence states them directly.
2
Evidence matching by population and condition prevents a real study from supporting the wrong claim.
Check the result
Does the ledger include implied, comparative, urgency, and guarantee claims?
Does each supported claim match evidence population, measure, condition, and timeframe?
Are release blockers separated from optional copy improvements?
ChatGPT