If I use a bot, will Meta block my WhatsApp number?

Using a bot is not grounds for a block. Meta doesn't penalize automation: on the official WhatsApp Cloud API, automating is the platform's intended use. What triggers blocks is the number's behavior: how many people block or report you, how many templates get rejected, and whether you write to people who never gave you their number. The risk changes depending on the route you take. With the Cloud API, a bad day lowers your messaging limit and it recovers within a week. With an unofficial solution based on Baileys (Evolution API, wppconnect, and most of the USD 20-a-month "WhatsApp APIs"), the number can disappear from one moment to the next, with no dashboard, no warning and no real appeal.

Meta doesn't look at your bot, it looks at how people react

The platform has no way of knowing whether there is a language model or a person typing on the other side. What it does measure, and publishes in WhatsApp Manager, is the number's quality rating: green (high), yellow (medium) or red (low). That rating is built from recipient feedback over a short rolling window —Meta documents seven days— and the signals that move it are three: people blocking you, people reporting you, and your templates piling up complaints.

From that comes a practical consequence: the quality of the bot's conversation is a compliance variable, not a comfort one. A bot that answers "I didn't understand" five times in a row and offers no way out to a human produces user blocks, and user blocks are exactly the signal Meta looks at. The handoff to a human in Chatwoot or in the CRM isn't there to make support friendlier: it's there so the number doesn't get burned.

The three blocks that get confused all the time

When someone says "they blocked my number" they may be talking about three different things, with very different severity and recovery times.

1. Low quality and a cut limit. The number shows up flagged in WhatsApp Manager and the messaging tier drops. The tiers are 250, 1,000, 10,000 and 100,000 unique recipients every 24 hours, plus the unlimited level. They go up on their own if you send volume with good quality, and they come down just as much on their own. There is no appeal because there is no penalty: it's a thermostat.

2. Number deactivated. The phone number is disabled inside the WhatsApp Business account. You can appeal from WhatsApp Manager. The rest of the numbers in the account keep working.

3. Business account disabled. It's the worst scenario and the one that costs the most money: the entire Business Manager goes down. If that same Business Manager holds your pixel, your catalogs and your Meta Ads campaigns, everything goes down together. That's why it's better for the WABA not to live in the same Business Manager as the advertising operation: separating them takes half an hour and prevents a WhatsApp problem from switching off your ad spend.

Cloud API and Baileys don't carry the same risk

The difference is not in the probability that something goes wrong, but in whether there is somewhere to find out beforehand and to file a claim afterwards.

With the official Cloud API (directly with Meta or through a BSP such as 360dialog or Twilio) the bot is a legitimate use. You have a visible quality rating, messaging tiers that go up on their own, templates with a status you can check, an appeal button and a history that follows you. The worst day leaves you restricted and recoverable.

With a Baileys-based solution —Evolution API is the most widely used in LATAM— your number ends up connected as just another WhatsApp Web linked device. That means you operate outside the Terms of Service, that there is no quality rating to monitor, that there is no dashboard where you can see things going badly before it's too late, and that the block can reach the number's WhatsApp account, not just the bot: the phone with that line stops working too.

That doesn't mean Evolution API is indefensible. For a clinic that answers 200 inbound queries a month and never broadcasts, the real risk is low and the savings are concrete. The problem appears when that same architecture is used to send: that same clinic firing off 300 appointment reminders a week is betting the line its patients already have saved, with no dashboard to warn it that things are going badly before the disconnection.

The five behaviors that actually trigger blocks

1. Cold sending to lists without opt-in. It's the number one cause, by far. You exported contacts from a spreadsheet, from the billing system or from a purchased database, and you sent. Meta doesn't publish the exact report threshold that pushes you to red; what we do know is that it's calculated over a few days, so a single bad afternoon is enough.
2. Miscategorized templates. Sending a promotional message under the utility category to pay less per message is the most common trick and the easiest to detect: Meta recategorizes it, and repeat offenses are paid for with rejections.
3. Templates with accumulated negative feedback. A template can move to paused status and stop being sent for a few hours, with longer and longer pauses if the pattern continues, until it ends up disabled. It's an early warning: if a template gets paused, the problem is the content or the audience, not the tool.
4. Forcing the 24-hour window. Outside that window, a free-form message fails with error 131047 and you have to use a template. So far that's technical. The problem is solving it by disguising commercial re-engagement as operational reminders.
5. Bots with no way out. No option to talk to a person and no explicit opt-out. One line is enough: "Reply STOP and we won't write to you again". It costs nothing and diverts to the cheap exit the people who would otherwise block you.

The same case, with numbers: 1,000 former gym members

Typical ranges for this kind of campaign, not audited results from a specific company.

The bad version. A marketing template is sent to all 1,000 contacts in one afternoon, from a recently enabled number with a tier of 1,000 every 24 hours. Cost at Meta: around USD 60, at a price on the order of USD 0.06 per marketing message in most LATAM countries (check the current rate for your country, Meta updates it). Between 4% and 8% block or report: that's 40 to 80 people. The rating falls to red within 24 to 48 hours and the tier is cut to 250. Typical recovery at a 2% to 4% response rate: between 20 and 40 conversations, with the number compromised for the next campaign.

The healthy version. You segment down to those who had some interaction in the last six months: 320 contacts. It's sent in batches of 80 per day over four days, with the explicit opt-out in the first line, and with the bot replying in under a minute to whoever answers. Cost at Meta: about USD 19 in templates, plus the service conversations that each reply opens, which within the 24-hour window are not charged. Blocks below 1%. Typical response of 15% to 25% on a warm list: between 48 and 80 real conversations.

A third of the spend, twice the conversations and the number moving up a tier instead of down. The difference wasn't the technology: both cases run on the same Cloud API with the same n8n behind them.

You've already been blocked: what to do in the first 48 hours

Before touching anything, identify which of the three cases it is. Go into WhatsApp Manager and look at the number's status.

• Low quality or a cut limit. Don't appeal, there is nothing to appeal. Stop 100% of proactive sending for seven days and only answer inbound conversations. The rolling window clears itself. Use those days to review which template caused the spike.
• Number deactivated. Appeal from WhatsApp Manager and wait between 24 and 72 hours. The generic appeal gets rejected. The one that works describes the specific use case, explains how you obtain consent and attaches the evidence: the screenshot of the web form with the acceptance checkbox, or the point in the flow where the customer asks to receive notifications.
• Business account disabled. The appeal goes through Meta's Accounts Center. It's the slowest one. In the meantime, don't open another business account with the same tax details, the same domain or the same card: Meta cross-references that information and the new account goes down too.
• Baileys or an unofficial solution. There is no dashboard and no process. If what got blocked is the number's WhatsApp account, the only channel is the support claim from within the app, and the recovery rate is low. The realistic plan B is a new number with a gradual volume ramp-up, and using the outage as an argument for migrating to the Cloud API.

In every case, what you must not do is retry in a loop or connect the number to another tool to see if it catches. That adds negative signal to a number that is already being watched.

Why the number has to be in your name and not the agency's

If you hire someone to build your bot, the WABA and the business account are created with your legal business name and your tax documentation, and the provider is added as a partner with permissions from the Business Manager settings. Never the other way around. Four concrete reasons:

• The number is a marketing asset. It's on your website, in your Google Business Profile, on the invoice and on your signage. Changing it costs more than the entire project.
• The history is not transferred. The quality rating and the messaging tier are built on that WABA. If you leave the provider and have to start over on another account, you go back to 250 recipients every 24 hours.
• Business verification goes with your paperwork. If the agency did it with theirs, the display name your customers see may end up being theirs.
• Moving a WABA between business accounts is possible, but it isn't one click. It requires both parties to enable it and Meta applies waiting windows. With the relationship already broken, that cooperation is exactly what you won't have.

It's the same logic as with your website's domain, with one aggravating factor: a domain is transferred with an authorization code and a week of patience, and a WABA in the hands of a provider you fell out with, isn't.

When ban risk isn't your problem

Three situations where this topic is oversold and you shouldn't spend on it:

• If your bot only replies to whoever writes first and you don't send a single proactive message, the real risk is marginal. Don't buy a premium BSP or "anti-ban protection". What you're missing isn't compliance: it's the bot answering well and booking appointments.
• If you have fewer than 50 conversations a month, migrating to the Cloud API is paperwork that doesn't pay off: business verification, display name, templates. Stick with the number handled manually or with a lightweight tool until the volume justifies the process.
• If your model is cold outbound over purchased lists, no architecture will save you. It's not a tool problem: Meta will switch off your number, and any provider promising you otherwise is selling you smoke. We don't take on that kind of project.

And if the same business account has already been blocked twice, the diagnosis isn't in the configuration. It's in who you're writing to.

This is exactly what we review in the 15-minute audit

In those 15 minutes we look at four things: whose name the number is under today, whether your sending is inbound or proactive, what share of your database gave you verifiable consent, and whether your volume justifies the Cloud API or not yet. From that comes a binary answer on whether it's worth changing anything or leaving the number as it is.

Write to us at info@striqtech.com or book the assessment at striqtech.com. If the diagnosis is that you don't need to migrate anything, we'll tell you that too.