What counts as spam on WhatsApp: what Meta punishes and how to sell without becoming a spammer

Everyone has gotten one: the "hi, how are you?" from an unknown number selling something. And every company that sells on WhatsApp has wondered where the line is — after all, what counts as spam on WhatsApp for Meta? The answer matters because the punishment isn't a slap on the wrist: it's a ladder that ends in a banned number. This guide shows what the algorithm punishes, how it detects it, and how to do active selling on the right side of the line.

The definition that matters: an unsolicited message

For WhatsApp, spam isn't a question of volume — it's about consent. The official rule from the Terms of Service: bulk, automated, or unsolicited messages are prohibited. In practice, the algorithm treats these as spam:

• A message to someone who never gave you their number (a purchased list, scraped from groups, a "referral" with no context)
• Abnormal volume of new messages in a short time, especially from a recently activated number
• The exact same identical text replicated to dozens of contacts
• Messages to people who don't have your number saved and don't reply — silence at scale is also a signal
• Automation layered on top of regular WhatsApp (QR Code bots)

How the algorithm detects it (without reading your messages)

Encryption stops Meta from reading the content — detection is behavioral:

• Report and block rate: the "Report" button on conversations with unknown senders is the main input. A few reports in a short time already raises a flag.
• Sending pattern: the ratio of sent to replied messages, speed, number of new recipients, the number's age.
• Device signals: modified apps (GB WhatsApp) and unofficial automations are detectable by client/version.

Translation: you can write the best message in the world — if you send it to 300 strangers at once, the pattern looks like a spammer's.

The punishment ladder

1. Reach drops/silent limits — messages to strangers show a "report and block" warning.
2. Temporary ban — the countdown screen ("try again in X hours"), common for automation and modified apps.
3. Permanent ban — "This account isn't allowed to use WhatsApp." The number dies, history and channel together — the review path is here.

Active selling without spam: the line in practice

Prospecting on WhatsApp isn't prohibited — spam is. The difference fits into four rules:

1. Consent first. Get the customer to initiate or authorize the contact: a wa.me link on the website and Instagram, a QR Code at the counter, a form with opt-in. Whoever starts the conversation never reports you.

2. Context in the first line. "Hi {{nome}}, you asked for a quote on our website yesterday" is legitimate contact; a dry "hi, how are you?" from an unknown number is scam behavior — and gets reported as such.

3. Scale only through the official path. Campaigns to your own base through the official API with approved templates: Meta knows it's a business, the limits are clear, and the penalty for a mistake is pausing the template, not killing the number.

4. Give an opt-out and honor it. "Reply STOP to stop receiving these" — and follow through. Whoever leaves through the door doesn't report you; whoever can't find the door does.

An organized business doesn't need spam

Spam usually comes out of operational desperation: with no pipeline, no history, and no follow-up, "blasting the whole list" looks like the only lever. With WhatsApp inside a CRM — every lead with a stage, automatic follow-up at the right time, AI replying in 24 seconds — revenue comes from working the base you've already earned. In our benchmark, the manual flow leaves the lead waiting 15 hours: the problem is rarely a lack of people to reach out, it's the hot lead going cold in the queue.

Frequently Asked Questions

Is sending a promotion to my customers spam? For real customers, with consent, a reasonable frequency, and an opt-out: no. It turns into spam when the base is cold or the frequency becomes a bombardment.

How many reports bring down a number? Meta doesn't publish the number — and it's relative to the account's volume and age. A new number with few reports goes down; an older number with a healthy base can take more. It's not a limit to "manage": it's a signal to bring to zero.

Can I message a number I got from a group? That's exactly the pattern the algorithm punishes: unsolicited contact. If you do, make it individual, with context from the shared group — and accept a no.

Is a broadcast list spam? Not by itself: it only delivers to people who have your number saved, which already filters for consent. The content and frequency are what decide.

Can spam cause legal trouble beyond a ban? It can — data protection regulations require a legal basis to use personal data for marketing. A purchased list is a double problem: it bans the number and exposes the company.