Rules for AI-written marketing claims: what a team should agree

The team's rules

1. List every claim a reader could check: numbers, rankings, comparisons, results, "only", "first", "guaranteed", "secure".
2. Attach the proof to each one, with its date and method. No proof, no claim.
3. Treat health, money, safety and compliance claims as high risk, and send them to whoever signs off legal points.
4. Use only real testimonials, with written permission and any payment or free product disclosed.
5. Put conditions next to the offer: "free" with "card required" in the same place.
6. Name one approver per piece, and keep the proof with the published version.

A worked example

Illustration: a made-up business. A 12-person company that sells scheduling software asks its AI for a landing page headline and three proof points. The marketing lead, Jun, runs the claims check on the draft:

Jun approves the page with the proof for each line saved beside it, and adds one line to the team's prompt: "Never write a testimonial, a number or a superlative we haven't given you."

Put the rules where the AI reads them

1. Add the never-list to the team's instructions: no invented testimonials, numbers or superlatives.
2. Give the AI your approved proof points as a file, so it uses those instead of inventing.
3. Run a claims check on every draft before review, not after publishing.
4. Keep customer data out of personal accounts; see what not to paste.
5. Have counsel read anything in health, finance, safety or comparison with a named competitor.

A summary of US rules, not legal advice. Other countries have their own advertising rules.

Next step: make the check a skill

A rule the team has to remember gets skipped on a busy day. Claude skills for marketing covers turning the claims check and your voice into skills the whole team's AI uses, and the AI usage policy template covers the company-wide rules.

Questions

Can we publish marketing claims that AI wrote?

Yes, if a person checks them against proof first. In the US, the FTC expects a reasonable basis for objective claims before they run, and a claim like 'tests prove' needs at least that level of proof. AI drafts often add superlatives and numbers nobody measured, so list every checkable claim and remove or source each one.

Can AI write customer reviews or testimonials?

No. The FTC's 2024 rule on reviews and testimonials covers reviews that misrepresent that they're by someone who doesn't exist, such as AI-generated fake reviews, and lets the agency seek civil penalties against knowing violators. Testimonials must come from real customers, with their permission.

What should a team's rules for AI-written marketing include?

Every checkable claim has proof on file before it runs; superlatives and comparisons need a current like-for-like test; testimonials are real and permitted; conditions of an offer sit next to it, not only in a footnote; and one named person approves each piece.