Before Publishing the Benefit: A 17-Point Claim Audit

Make the audit a release gate, not a late copy edit

A claim audit begins before the benefit appears in a campaign, sales page, email, demonstration, testimonial, comparison, or affiliate placement. Freeze the exact version under review and name the owner who can stop release. The artifact should connect the likely message to evidence, conditions, disclosures, approvals, and every downstream variant. A general folder of research is not enough when nobody can trace which sentence it supports.

This checklist explains selected United States guidance and independent policy research for education. It cannot decide compliance in another jurisdiction or replace qualified legal and subject-matter review. Health, medical, financial, or safety claims require appropriate specialists. Mark an item pass only when a reviewer can inspect the record; uncertainty is review or stop, never an assumed yes.

Evidence: U.S. Federal Trade Commission; U.S. Federal Trade Commission; OECD

Points 1–4: identify the message and the thing being sold

First, capture the exact words, images, audio, demonstration, placement, call to action, and surrounding page. Second, ask an uninvolved reviewer to write the strongest reasonable express and implied messages. Third, identify the precise product, service, plan, formulation, configuration, region, and version. Fourth, record the audience and the decision the message is likely to influence.

These fields prevent a common category error: evidence for one feature is attached to a broader promise about another version or user. Review the mobile view, shortened ad, metadata, and creative crop because each may produce a different overall impression. Preserve the reviewed artifact or a stable snapshot where lawful and proportionate.

Evidence: U.S. Federal Trade Commission; U.S. Federal Trade Commission

Points 5–9: test evidence fit rather than evidence volume

Fifth, classify the assertion as feature, mechanism, performance, comparison, typical result, health, financial, environmental, safety, or another regulated type. Sixth, record evidence that existed before dissemination. Seventh, match its population and setting to the audience. Eighth, match the measured outcome, baseline, duration, and comparison to the advertised result. Ninth, document uncertainty, limitations, contradictions, and missing observations.

A stack of links does not cure a mismatch. A mechanism does not establish a typical outcome; an average can hide a wide distribution; a relative improvement without its baseline can distort practical significance. Record who produced the evidence, what method was used, and whether an independent source corroborates the interpretation. Escalate claims of scientific proof or consequential outcomes to qualified reviewers.

Evidence: U.S. Federal Trade Commission; U.S. Federal Trade Commission

Points 10–13: expose typicality, conditions, and relationships

Tenth, determine what ordinary users should expect, not merely what happened in a selected success. Eleventh, list conditions that materially change the result: training, support, frequency, compatible equipment, extra purchases, exclusions, or uncommon effort. Twelfth, place necessary qualifications where they can affect the decision. Thirteenth, reveal endorsements, affiliate compensation, sponsorship, sampling, or another commercial relationship in understandable language.

Evaluate the net impression after these additions. A tiny qualifier cannot neutralize an unbounded headline, and a disclosure that arrives after a purchase decision is not useful decision support. Test proximity, prominence, language, scrolling, interaction, audio duration, screen size, and accessibility. If accurate qualification makes the main promise collapse, rewrite or remove the promise.

Evidence: U.S. Federal Trade Commission; U.S. Federal Trade Commission; OECD

Points 14–17: control reuse, approval, correction, and scope

Fourteenth, enumerate every reuse location: landing pages, ads, email, social posts, video, marketplace text, partner kits, translations, structured data, and cached variants. Fifteenth, record named editorial, evidence, legal, and specialist approvals required for this category. Sixteenth, assign a correction owner, withdrawal route, and trigger such as product change, new research, complaint pattern, or expired evidence. Seventeenth, identify the jurisdictions, channels, and dates actually reviewed.

Do not approve a master sentence while ignoring a cropped derivative that removes its limits. The propagation list should let the owner locate and correct every material instance. Keep rejected copy and the reason for rejection so a future campaign does not revive it without review. If ownership or removal access is missing, release remains blocked.

Evidence: U.S. Federal Trade Commission; U.S. Federal Trade Commission

Record pass, revise, stop, and the next review date

Use four outcomes instead of a vague quality score. Pass means the bounded message and evidence align. Revise means wording or presentation can be corrected without changing the evidence. Specialist review means the subject or jurisdiction exceeds the team's competence. Stop means support is absent, a material omission remains, or the message creates unacceptable rights or safety risk. Never average a stop condition into an otherwise favorable total.

Before release, read the page once as a skeptical buyer with none of the team's background knowledge. Then read it as the correction owner who must defend every element six months later. Recheck no later than 2027-02-10 and sooner when product, audience, evidence, interface, price, commercial relationship, or applicable rules change. OECD work on dark commercial patterns is used here to broaden the choice-architecture review, not to pronounce a legal violation.

• Pass only for the reviewed version and scope.
• Revise copy when the evidence supports a narrower message.
• Escalate consequential or regulated subjects to specialists.
• Stop publication when substantiation or correction control is missing.

Evidence: U.S. Federal Trade Commission; U.S. Federal Trade Commission; OECD