Method

How MESSAGE works

What MESSAGE does, what the AI sees, how the seven parts work, and what to keep in mind.

What is MESSAGE

MESSAGE helps pharmaceutical marketing teams think through a message before it goes further. It asks one question: are we saying the right thing, to the right audience, in the right way?

It is not a proofreader and it is not a compliance checker. It does not approve anything. It gives your team something to react to.

How it works

  1. Paste and describe. Paste the message and say what it is for: the product, the audience and the key benefits. Optionally add your evidence, a competitor message and what the audience believes today. No message yet? Tick the box and MESSAGE writes one.
  2. Read the result. Up to about 45 checks in seven parts, each green, yellow or red, with a plain explanation, a suggested rewording and the framework it rests on. Checks that do not fit your case are listed as not run, with the reason. If you added evidence, each claim in the message is shown as backed or not backed.
  3. Use the improved version. MESSAGE gives an improved version with the fixes applied. You can use it as your message and check it again, or rewrite it for healthcare professionals or for patients and caregivers.

The seven parts

The checks you get depend on your audience, channel and what you pasted. Each one says what it rests on: a published instrument, a regulation, a practitioner convention, or MESSAGE's own check.

  • 1. Message strategy. One main idea, in the right place, with supports that earn it (message house, benefit ladder, CDC Clear Communication Index).
  • 2. Audience fit. Plain words, active voice, scannable layout, and a clear next step for the public; data first and clinical language for professionals; the questions payers ask.
  • 3. Behaviour and emotion. Where the reader is now and what moves them: benefit, barriers, susceptibility, confidence, cues, stage of change (Health Belief Model, COM-B, Stages of Change).
  • 4. Differentiation. A specific target, a frame of reference, a point of difference and a reason to believe (positioning template). Stock phrases and overlap with the competitor line are counted in code.
  • 5. Evidence and claims. Endpoint, surrogate, population and significance against the evidence you pasted. Numbers in the message that are not in your evidence are flagged in code.
  • 6. Risk-language signals. Balance, prominence, superiority, indication and omitted material. These are prompts for your team, not a compliance verdict.
  • 7. Language and craft. Reading grade, sentence length, channel length, hedging and shouting, all counted in code so the answer is the same every time.

Your data

Your project is saved in this browser only. Clearing the browser's site data removes it. Nothing is stored on a server.

When you build, rewrite or check a message, the text you entered is sent to an AI service run by Cloudflare (Workers AI, free allowance) and the answer comes back. MESSAGE does not keep it. Read Cloudflare's data-usage terms before you paste unreleased messaging, and never paste patient information.

Limits

  • The AI can be wrong, and it can miss things. Treat every answer as a prompt for discussion.
  • MESSAGE can only judge against the evidence you paste. It cannot tell whether that evidence is the approved version.
  • The free AI has a daily allowance. When it is used up, MESSAGE says so and works again after midnight UTC.