Is your AI bot allowed on the WhatsApp Business API?
Under Meta's 2026 terms a support bot is fine and a general-purpose assistant is not. A test with examples, the gray areas, and what happens if Meta rules you out.

Key takeaways
- The test in Meta's terms is whether the AI is the primary, rather than incidental or ancillary, thing you offer, and Meta decides in its sole discretion.
- Order support, booking and FAQ bots for your own business are allowed; companions, personal assistants and ask-me-anything bots are likely barred.
- The answer changes by user country: regulators have forced Meta to allow general-purpose assistants for users in the EEA, Italy and Brazil.
- Meta's published enforcement ladder runs from warnings to temporary blocks to permanent disablement of a number or account, with appeals usually decided in 24 to 48 hours.
It depends on what the AI is for and where your users are. A bot that serves your own customers (order support, bookings, FAQs) is allowed. A bot whose product is the AI itself, such as an assistant, companion or “ask me anything” chat, has been barred since 15 January 2026, except for users in the EEA, Italy and Brazil, where regulators forced Meta to allow it.
What does Meta’s rule test?
Meta’s AI Providers clause tests one thing: whether AI technology is the primary, rather than incidental or ancillary, functionality you make available. Meta’s WhatsApp Business Solution Terms bar providers and developers of AI technology, including general-purpose AI assistants, from the WhatsApp Business Platform in that case. Meta decides what counts, in its sole discretion. The Italian competition authority quotes the full clause in its December 2025 decision, and Meta publishes the terms on its Business Solution Terms page.
Because Meta decides case by case, there is no objective test you can pass in advance. The best stand-in is a question: what did the person messaging you come for? If they came for your business’s service (a parcel, a table, an appointment) and AI does the work, the AI is ancillary. If they came to talk to the AI, the AI is the primary thing you offer.
The model behind the bot is not the test. Meta says businesses that use AI to serve their own customers are not affected (TechCrunch), and the European Commission describes the allowed side as AI used for ancillary or support functions, such as automated customer support.
How do common bots score?
Seven kinds of bot cover most of what people build. The verdicts are our reading of the clause and of what Meta, the Commission and the AGCM have said. We found no list of examples published by Meta.
| Bot | Verdict | Why |
|---|---|---|
| Order, delivery or support bot for your own business | Allowed | The AI serves your own customers. Meta and the Commission both name this use as allowed. |
| Booking or reservation agent for your own business | Allowed | The AI does a business task, and the booking is the service. |
| Helpdesk for a school or course (enrollment, schedule, billing) | Allowed | The AI answers about your service. It is not the service. |
| AI tutor sold as the product | Likely barred | The conversation with the AI is what the customer buys. |
| AI companion | Likely barred | Nothing else is delivered. The AI is the product. |
| Personal assistant or “ask me anything” bot | Barred | These are the general-purpose assistants the clause names. |
| Concierge that books for many businesses | Gray | It serves its own users, not one business’s customers. The answer turns on whether the AI or the booking is the product. |
Does your users’ country change the answer?
Yes, because Meta’s treatment of AI providers follows the country of the user receiving the message. Meta’s pricing page for AI providers says general-purpose assistants are permitted only where Meta is “legally required to permit this use case”. As of 19 October 2026, users in the EEA (including Italy) can receive them at no charge, Brazil allows them with a per-message charge, and Meta’s page added Türkiye on 15 October 2026 without stating a legal basis. Everywhere else the clause applies as written.
The EEA access rests on interim measures from the European Commission dated 9 June 2026, which Meta is appealing. A bot that depends on it can lose access without any change in your product. WhatsApp’s AI chatbot ban: the rule and where it stands has the dates and the status of each case.
Where are the gray areas?
Four situations are hard to classify, and Meta’s discretion makes each of them a judgment call.
- A support bot that will chat about anything. If its prompt lets it answer any topic, the AI starts to look like the main thing. Keep its scope to your own business and give people a path to a human. WhatsApp’s Business Messaging Policy allows automation inside the 24-hour customer service window and expects a clear way to reach a person.
- A paid conversation with an AI inside a business wrapper. A tutor, coach or expert persona that charges for the chat is selling the AI. A school that answers questions about its courses is not.
- Resellers and platforms. The clause bars access “directly or indirectly”. On a plain reading, routing a general-purpose assistant through a platform or Business Solution Provider does not change the answer.
- Your own model. If you develop an LLM and the product is chatting with it, you are the kind of “AI Provider” the clause describes, however small the audience.
WhatsApp chatbot vs AI agent covers how the two differ and why the rule treats them differently.
What happens if Meta decides you are out?
Meta’s Policy Enforcement page describes an escalating ladder for breaches of the Business Messaging Policy, Commerce Policy and Business Terms. It starts with a warning. Next come blocks on sending: 1 or 3 days for templates, then 5, 7 or 30 days for all messages. After that comes an account lock, which lasts until an appeal succeeds, and finally permanent disablement of the phone number or account. Evidence of severe harm skips the ladder.
You can appeal in Business Support Home, and the page says decisions typically take 24 to 48 hours. WhatsApp’s Business policy adds that it may bar an organization from all future use of its products after a termination.
That ladder is written for violations in general. We found no published case where Meta enforced the AI Providers clause, and no description of how it detects a breach. If your bot sits near the line, do not make WhatsApp its only channel.
What can you do if your bot does not pass?
A bot that fails the test has four options, and they are not exclusive.
- Narrow the product. Make the AI serve a business function and drop the open-ended chat. This is the one change that moves a bot from barred to allowed everywhere.
- Serve EEA users on the Business API. It works today, and it is exposed to the appeal.
- Move the AI conversation off the Business API. Your own app, the web or Telegram carry no WhatsApp rule. Put an AI agent on WhatsApp: every option compares all the routes.
- Use a personal number. With an open-source library on your own number, the ban risk is yours. With whasp, you get a real personal WhatsApp account on a cloud phone that whasp runs: people text it first, every message reaches your webhook, and one API call replies. See pricing and the docs.
A personal account is not a Business Platform account, so the AI Providers clause is not the rule that governs it. Personal accounts follow WhatsApp’s general Terms of Service, and numbers that behave like bots, above all ones that message strangers, get banned: that risk is real. whasp is built around it (users write first, paced sending, its own IP per number) and replaces a banned number at no cost, though chats do not move to the new one. If you are building a companion, how to launch an AI companion on WhatsApp covers the full path.
Questions
Can ChatGPT run on the WhatsApp Business API?
As the model inside your own support bot, yes: the model behind a bot is not the test. As OpenAI's own ChatGPT assistant, it was barred from 15 January 2026 and has returned only for users in the EEA, since 13 July 2026.
Is an AI companion allowed on the WhatsApp Business API?
Probably not. A companion's product is the AI conversation itself, which is the primary functionality the rule targets. Users in the EEA, Italy and Brazil are the exception, because regulators forced Meta to allow it.
Can Meta ban my WhatsApp Business number for running an AI bot?
Meta can restrict or permanently disable accounts that breach the Business Terms. We found no published case of enforcement under the AI Providers clause, so how Meta applies it in practice is not documented.
Does the rule apply to a personal WhatsApp number?
The clause is in the Business Solution Terms, which govern the Business Platform. A personal account follows WhatsApp's general Terms of Service instead, where the risk is a ban when a number behaves like a bot.


