> ## Documentation Index
> Fetch the complete documentation index at: https://howto.paigeme.dev/llms.txt
> Use this file to discover all available pages before exploring further.

# Plan broadcasts by chatting with the Broadcasting assistant

> The Broadcasting assistant builds audiences, drafts templates, and assembles whole campaigns from a conversation — then hands them to you to approve. It can never send anything itself.

Open the **Broadcasts** tab and the chat sidebar switches to the **Broadcasting** assistant. It's a different agent from the one that writes your bot code, and it knows about your contacts, your segments, your templates, and your past broadcasts.

It's most useful for the things the forms make tedious:

> "How many opted-in contacts do we have tagged vip?"

> "Make a segment of everyone who tapped Book now in the last 30 days."

> "Draft a utility template telling people we're closed on the 24th."

> "Put together a campaign for that segment using the closed\_notice template, for Friday at 9am."

## What it can do

* **Answer questions** about your broadcasts, segments, templates, and audience sizes.
* **Check an audience size** before you commit to anything — the number it gives is the same one the send would use.
* **Create and edit segments**, including targeting by **buttons people tapped**, which the Segments form can't do.
* **Draft, submit, and check templates.** It'll compose one, show it to you, and only submit to Meta once you say so. It can then tell you whether Meta has approved it yet.
* **Assemble a campaign** — audience, template, variables, schedule — and hand it to you.

## It never sends anything

<Note>
  The assistant cannot send, launch, approve, cancel, or retry a broadcast. There's no tool for it, and the system refuses to send a campaign in the state the assistant leaves it in. Asking it to "just send it" won't work — it'll explain that approving is yours to do.
</Note>

That's deliberate. An agent that could send a promotion to four thousand people because it misread a sentence is not a good agent.

## The approval card

When you ask it to assemble a campaign, a card appears in the chat with everything laid out: the name, how many recipients, the estimated cost, the schedule, a preview of the template as your customers will see it, and each variable with what it resolves to.

Three buttons:

* **Approve** — you're happy. The campaign is scheduled and sends at its time.
* **Edit** — tell the assistant what to change and it reassembles.
* **Reject** — discard it.

<Warning>
  **Approve means it will send.** If the campaign has no schedule, or a time that's already passed, approving sends it at the next opportunity — within minutes. The wizard's rule that nothing sends without a schedule doesn't save you here; your Approve click *is* the decision. Check the schedule line on the card before you click.
</Warning>

If the campaign uses a template Meta hasn't approved yet, approving is still fine — the campaign waits, and sends by itself once Meta approves and the scheduled time arrives. If Meta rejects the template instead, the campaign is cancelled and you'll get a **Campaign needs attention** notification.

## Marketing templates and opt-in

The assistant refuses to assemble a **marketing** campaign to an audience that isn't restricted to opted-in contacts, and tells you to fix the audience. Marketing messages to people who never opted in are exactly the thing that gets a WhatsApp number restricted.

<Note>
  The manual wizard doesn't enforce this — you can untick **Opted in only** and pick a marketing template there. The assistant is stricter than the form on purpose. Being stopped here is a reason to reconsider the audience, not to go and do it by hand.
</Note>

It also won't assemble a campaign with no audience at all, rather than quietly targeting every contact you have.