Skip to content
Search
AppliedPlaybooks

Brand Voice Prompt and a Claims Check Before You Publish

Difficulty
Beginner
Time
1 hour to build, 5 minutes per check

Before you start

  • Five to ten pieces of copy your company has already approved and published (web pages, emails, posts)
  • The proof behind your main claims: test results, customer counts, certifications, reviews
  • An AI chat assistant your company has approved for marketing work

This brand voice prompt turns your best approved copy into a voice guide an AI assistant can follow, and adds a claims check you run before anything is published. A generic instruction like "friendly but professional" gives generic copy. Rules drawn from your own sentences give copy that sounds like you. The claims check matters for a different reason. In the United States, the Federal Trade Commission expects advertisers to have a reasonable basis for their claims before they publish them, and an AI draft can add a claim nobody at your company ever made.

Step 1: Collect the approved copy

Use the AI assistant your company has approved for this kind of work, and follow your company's AI policy. If you don't know which tool is approved, ask before you paste anything. Use copy that is already public. Don't paste unreleased plans, customer lists or anything marked confidential.

Step 2: Draft the voice guide from quotes

Below are pieces of our approved, published copy. Write a voice guide
from them.

For each rule (for example "short sentences", "we say 'you', not
'customers'"), quote one sentence from the copy that shows it.
If you can't quote an example, leave the rule out.

Then list:
- Words and phrases we use often.
- Words and phrases we never use (only ones absent from all samples
  that a generic draft would likely use).
- What we sound like in one sentence.

Copy:
[paste]

Read every rule and its quote. Delete any rule where the quote doesn't actually show it. Add the phrases your team already bans, even if the AI didn't find them.

Step 3: Build the approved-claims list

Write the claims you're allowed to make, each with its proof and where the proof lives. Two made-up examples of the format: "Rated 4.8 by 312 reviewers [link to review page]" and "ISO/IEC 27001 certified [certificate on file]". This list is written by people, not the AI. If you can't name the proof, the claim doesn't go on the list.

Step 4: Use both when you draft

Put the guide and the claims list at the top of any drafting chat, or in your assistant's saved instructions if it has them (that's what a system prompt is for). Add: "Only make claims from the approved list. If a claim would help but isn't on the list, write [CLAIM NEEDED] instead."

Step 5: Run the claims check before publishing

Below is a draft and our approved-claims list. List every claim in the
draft: any statement about results, numbers, speed, safety, rankings,
awards, certifications, prices, comparisons or what the product does.
Mark each: APPROVED (quote the list line), NOT ON LIST, or CHANGED
(on the list but worded more strongly than the list allows).

Draft:
[paste]

Approved claims:
[paste]

Fix or remove every NOT ON LIST and CHANGED line. Then read the draft yourself, because the check can miss a claim. Anything health, legal or financial in nature goes to whoever signs those off at your company.

What to do next

Keep the voice guide and claims list in one shared place and give each an owner. Add a claim only when its proof is on file. Run the Step 5 check on every piece before it's scheduled. If your team doesn't have rules yet for what goes into AI tools, start with the AI use policy playbook.

Rather have someone set it up for you?

That's a fair call. LTFI, the agentic business operating system from Kief Studio, can set up the work these guides describe and run it with you. No website yet? Kief Studio will build you one for free.

Go further on kief.studio