Manage WhatsApp health, opt-outs, and pause your AI on Botline

Two numbers from Meta decide what your WhatsApp Cloud API number can do tomorrow: Quality Rating and Messaging Tier. Both update silently in Meta's backend; both can quietly throttle your reach without ever sending you a notification email. Botline now surfaces them in real time on a dedicated health page so you stop finding out about a degradation when a broadcast underperforms.

Quality Rating is Meta's rolling read on how customers feel about messages from your number. It's computed from block rate, report rate, and 24-hour reply rate. If too many people block you or report you as spam, you slide from Green to Yellow to Red — and Meta starts limiting how many recipients you can reach in a 24-hour window.

Messaging Tier is how many unique customers you're allowed to start a conversation with in any 24-hour window. New numbers start at TIER_50 (50 unique recipients per day). With clean behaviour and a Green rating, you climb automatically to TIER_250, TIER_1K, TIER_10K, and finally TIER_100K. There's no application form — Meta just bumps you when the data says you're ready.

If you're running broadcasts at any scale, both of these matter on Monday morning when you ship a campaign. The health page gives you both at a glance, plus the controls to act on them.

Open Dashboard → WhatsApp → Health. The page lists every Cloud API line on your account with three live data points per line:

• Quality Rating chip — Green, Yellow, or Red. Updates in real time from Meta's webhook.
• Messaging Tier badge — TIER_50, TIER_250, TIER_1K, TIER_10K, or TIER_100K, with the unique-recipient cap spelled out (e.g. "up to 1,000 unique recipients per 24h").
• Opt-out count — how many customers have opted out of your broadcasts to date, with a link into the registry.

Plus a single big Pause / Resume button per line, and a Manual opt-out add field for when a customer asks you in person, by email, or on a call. The page is read-mostly — most days you'll glance at it and move on. The days it matters, it matters a lot.

The three colors map directly to what you can and can't do:

• Green — high quality. No restrictions. Keep doing what you're doing. Most well-behaved Botline tenants never leave Green.
• Yellow — degraded. Meta is flagging higher-than-normal block or report rates from your number. Marketing broadcasts still send, but you'll see a soft warning banner in the broadcast composer asking you to confirm. Treat this as an early-warning signal: your last 1-2 broadcasts touched a nerve. Pause new broadcasts, review your audience, tighten your opt-in flow before sending again.
• Red — low quality. Meta has slowed delivery on your number and you're close to a tier downgrade or temporary suspension. The broadcast composer adds a hard soft-block (with a one-click override that we strongly recommend you don't click). If you keep broadcasting through Red, the next step is a downgrade from TIER_1K to TIER_250 — or a 24-72h send freeze.

Recovery is mechanical, not negotiated:

• Stop sending marketing-tier templates for 24-72 hours.
• Reply normally to inbound messages — utility templates and 24-hour-window replies are fine and actively help your rating recover.
• Audit your last 100 sends. Anyone who didn't opt in clearly? Add them to the opt-out list manually before your next broadcast.

A Yellow number usually returns to Green within 24 hours of clean behaviour. A Red number can take 48-72 hours. There's no escalation path — Meta doesn't answer support requests on Quality Rating.

Your tier is the hard cap on how many new conversations (unique recipients) you can start in any rolling 24-hour window. Replies inside the 24-hour service window don't count. The five tiers Meta uses today:

• TIER_50 — up to 50 unique recipients per 24h. Default for new numbers.
• TIER_250 — up to 250 unique recipients per 24h.
• TIER_1K — up to 1,000 unique recipients per 24h. Most growing tenants land here.
• TIER_10K — up to 10,000 unique recipients per 24h.
• TIER_100K — up to 100,000 unique recipients per 24h. Top tier; effectively unlimited for most B2C use cases.

You climb automatically. Meta evaluates your number every 24 hours; if you're Green, sending consistent volume, and not getting reports, the next bump usually happens within 7-30 days. The fastest path up:

• Get explicit opt-in. A tap on a "Yes, send me updates" button or a checkbox at checkout is the single biggest signal Meta watches.
• Send Utility templates first. Order updates, appointment reminders, and account notifications have ~zero block rate and pull your reputation up.
• Reply quickly to inbound. 24-hour-window reply rate is a tier signal too — it tells Meta this is a real two-way line, not a one-way broadcast cannon.

You can't buy your way up a tier. You can only behave your way up. The good news: once you're at TIER_10K or TIER_100K with a Green rating, you stay there as long as your behaviour stays clean.

This is the section to read if you skip the rest. WhatsApp compliance hinges on respecting opt-out keywords — and Botline now does this automatically for every Cloud API line.

Automatic detection. When a customer replies to your number with any of STOP, UNSUBSCRIBE, END, QUIT, CANCEL, or STOPALL (case-insensitive, exact match), three things happen in the same transaction:

1. The customer's phone number is added to your opt-out registry with a timestamp and the keyword they used.
2. A short compliance confirmation reply is sent automatically (e.g. "You've been unsubscribed. We won't send you marketing messages. Reply START to opt back in.").
3. The AI is skipped for that turn — no automatic response, no draft, nothing else queues.

Broadcasts auto-skip opted-out contacts. When you build a broadcast audience, the composer automatically excludes every number on your opt-out registry — no checkbox to forget, no toggle to flip. The audience preview shows the final count after exclusions, so the cost preflight you see is the cost you pay. This is the single most important guardrail for tenants running marketing broadcasts at scale: you cannot accidentally re-spam someone who told you to stop.

Manual opt-out add. Customers don't always reply STOP. Sometimes they call, send an email, or tell you in person. Use the Add manual opt-out field on the health page to add their number with a short note. They'll be excluded from broadcasts the same as keyword-detected opt-outs.

Restore. Mistakes happen — wrong number entered, customer changed their mind by phone. Click Restore next to any registry entry to remove them. The action is audited and restoration takes effect on the next broadcast.

Opt-out isn't one-way. When a previously opted-out customer replies with any of START, OPTIN, OPT IN, RESTART, or SUBSCRIBE, Botline automatically:

1. Removes them from the opt-out registry.
2. Sends a welcome-back compliance reply (e.g. "Welcome back — you're re-subscribed to our updates.").
3. Resumes normal AI handling on the next message.

This is part of Meta's consent model: a customer who explicitly opts back in must be re-included automatically. Botline handles this for you so you never have to triage re-opt-in requests by hand.

Some days you want the AI off this line, but you don't want to disconnect the number or lose any inbound messages. The Pause button on the health page is for those days.

When to pause:

• Incident. The AI is misbehaving for some reason — bad answer style, hallucinating availability, replying in the wrong language — and you need a hard stop while you investigate.
• Vacation or holiday. Owner is offline for a few days and you'd rather customers see a polite "we'll get back tomorrow" than an AI that doesn't know about an out-of-office.
• Deploying changes. You're editing the system prompt, swapping the model, or rebuilding the knowledge base. Pause the line, ship the change, smoke-test in the conversations page yourself, then resume.
• AI-misbehaving emergency. Something is genuinely going wrong (we've all had a day). Pause first, debug second.

What pause does:

• Customer messages still queue — every inbound message is logged and visible in the conversations page exactly as before. Nothing is dropped.
• The AI does not auto-reply for the duration of the pause. No drafts queue either.
• Broadcasts skip the line — any scheduled or in-flight broadcast that targets this line is paused along with it.
• You can still reply manually from the conversations page. Pause is about the AI, not about you.

Resuming. Click Resume on the health page. The AI picks up from the next inbound message (it won't auto-catch-up on messages received during the pause — that's by design, to avoid a deluge of stale replies).

The two most common health-related questions:

• "Customer says they sent STOP but they're still getting messages." Check the broadcast schedule. Broadcasts queued before the customer's STOP message have already locked in their audience snapshot — Meta will deliver those messages even after the customer opts out. Botline excludes opted-out contacts from future broadcasts, not from in-flight ones. The conservative habit: when you detect an opt-out via the health page registry, manually cancel any pending broadcasts that targeted that customer if it matters. Going forward, the opt-out is permanent until they reply START.
• "My Quality Rating just dropped to Red — what now?" Open the broadcast composer. You'll see a cost-saver banner blocking new marketing-tier sends. Don't override. Open the health page, audit your opt-out registry — if the count jumped recently, your last broadcast hit an audience that didn't want to be there. Pause the line for 24-48 hours, focus on inbound replies and Utility templates only, and the rating typically recovers automatically.

For everything else — provisioning a new Cloud API line, climbing tiers, or wiring opt-in flows into your booking page or Shopify checkout — see the Cloud API guide and the Broadcasts & Templates walkthrough.

Health monitoring composes naturally with the rest of Botline:

• Pair the health page with the broadcast composer — the composer reads the live Quality Rating before every send and shows the right guardrail.
• Use Pause as your safety net before any big change: prompt edits, model swaps, knowledge-base rebuilds, or shipping a new catalog drop. Pause, change, smoke-test, resume.
• If you haven't connected Cloud API yet, the health page only lights up for Cloud API lines — see the connection guide first.

Open Health, glance at your rating and tier, and let the STOP-keyword detector keep your compliance clean while you focus on growth.