Workflow · Marketing and communications

Draft marketing materials with AI and check every claim

More options to choose from, with no claim you can't support.

Try with checksIt is worth trying only with the checks on this page.

For: Marketers, and small-business owners doing their own marketing

Go to the steps Test it on your own work

What the evidence says

Studied directly. Studies look at this task or something very close to it.

AI-written marketing content: not enough evidence yet (evidence strength 0 of 5)

Too few strong studies to rate. Most test whether AI copy performs, not the cost of checking it.

This workflow itself is our recommended way of working, not a tested method. Studies searched to 2 November 2026.

Before you start

Collect the facts you may use: product details, prices, approved claims and your tone of voice.
Check the advertising rules in your country and sector. You are responsible for what you publish, whoever or whatever wrote it.

The steps

Replace the words in square brackets with your own. Stay in the same chat for the next prompt, unless a step says otherwise.

You
Write the audience, the one action you want and the claims you are allowed to make.

Ask for different angles, using only your approved facts.

Prompt

Write 5 versions of [ad, email or post] for [audience]. The one action I want: [action].

Use only these facts and claims:
[paste]

Do not invent numbers, customer quotes, awards or comparisons with competitors. Each version should use a different angle, and name that angle. Keep to [length] and copy only the style of these examples, not their claims: [examples of our writing]. Don't use words such as best or freshest unless my facts prove them.

You
Choose one or two. Rewrite them until they sound like you.

Ask for a list of every claim in the version you chose.

Prompt

List every factual claim in this text, including implied ones such as 'best', 'fastest', 'trusted' or 'guaranteed'. Mark each one as: in my approved facts, stretched beyond them, or not in them. Do not defend them.

My approved facts:
[paste]

Text:

[paste]

You
Remove or prove each unsupported claim. Then test it against your usual version if you can.

How to catch its mistakes

Each claim is on your approved list or has evidence you could show a regulator.
No invented testimonials, statistics or reviews. Publishing them can be illegal as well as embarrassing.
No competitor is named or compared without evidence.
Images of people or places are disclosed where the rules, or your audience, expect it.

When not to use this

Health, financial or children's products, unless someone who knows those rules signs it off.
Publishing automatically with no person reading it first.

Tools

A general AI assistant. Examples: ChatGPT, Gemini, Microsoft Copilot. We chose these because they are common in workplace software, not because they are the best. We leave out Claude products as examples, because we draft this site with Claude. AI assistants compared has them all, with dated facts. Use whichever your employer has approved. The steps work with any tool of this type. We take no money from AI companies.

Find out if it works for you

People often misjudge how much AI helps them. Try this on your next 10 similar tasks, letting a coin flip decide which ones use AI. What to measure: The result that matters, such as click or sign-up rate, against your usual version. Your results stay in your browser. Start the self-test →

If you manage people

Keep one approved-claims list. Judge by results against a control, not by how much content was produced.

Written by AI Adoption's editors as our own recommendation, last checked 5 November 2026. It has not been reviewed by an outside practitioner. Before publishing, we ran its prompts once on made-up samples with a second AI model, also from Anthropic. That is a check on the wording, not a test of whether the workflow works for you. Disclosure: AI Adoption uses AI tools in its research and editorial work, including Claude from Anthropic. No company has any say in this page. Reuse under CC BY 4.0. Spotted an error? Tell us.