What we will not generate

Generation runs through a content guard. Most people never meet it; when you do, this page explains what happened.

What it refuses

Broadly, content that would put you or your audience at risk:

• Claims that would be unlawful or dangerous in a regulated field: medical, financial, legal
• Content that targets or demeans people
• Content that impersonates a real person or organisation
• Material that would breach the platforms' own rules

Why the line sits where it does

The product publishes under your name, to your audience, often automatically. A generated sentence that overstates a medical outcome is not an embarrassing draft: it is a claim your business made in public.

So the guard is deliberately conservative where the consequences are asymmetric, and that means it will occasionally refuse something you could have written safely yourself.

Regulated industries

Businesses in medical and similar fields can use the product, and the profile supports being specific about what you do. What the product will not do is write claims about outcomes, safety or efficacy on your behalf.

The practical consequence: it will help you explain a service, answer a common question and describe a process. It will not tell your audience that something works.

When it blocks something legitimate

It happens, usually because a topic is phrased in a way that resembles a claim. Try:

1. Rewriting the topic to describe the subject rather than the promise. "How our consultation works" rather than "Why our treatment works".
2. Being more specific. Vague phrasing about results reads as a claim; a concrete process does not.
3. Writing it yourself at Posts → New post. Your own words are yours to publish.

Automated messages are separate

No automated message is composed by a model at send time: every message an automation sends is text a person wrote. That is a stronger guarantee than filtering, and the reasoning is in automations overview.

You are still the publisher

The guard reduces risk; it does not transfer responsibility. What goes out under your name is yours, including anything generated. Read it before it publishes, which is what client approvals exists to formalise when someone else's name is on it.