Is AI allowed on WhatsApp Business?
Yes, if it serves your own customers. From 15 January 2026 Meta bars general-purpose AI chatbots from the WhatsApp Business API, not the assistants businesses use to look after their customers. Here is what changed, when, and how to keep an assistant inside the rules.
The short answer
No, a customer service assistant is not banned. If your business uses AI on WhatsApp to answer its own customers about its own services, bookings or orders, you are not the target of the 2026 rule. TechCrunch put it simply: businesses using AI for customer service are still allowed, and gave the example of a travel company's support bot.
What Meta banned are general-purpose AI chatbots: open-domain assistants that answer anything, offered by AI companies through the WhatsApp Business API. respond.io names the model directly: ChatGPT on WhatsApp, or Perplexity's conversational bot.
The line is drawn by purpose. Meta's WhatsApp Business Solution Terms stop AI providers from using the platform to offer their AI when it is “the primary (rather than incidental or ancillary) functionality”. A support assistant exists to serve your customers. The AI is how it works, not what you are selling.
Allowed: an assistant that looks after your customers. Not allowed: an AI assistant that is itself the product.
What changed and when
The change came in two steps, first for new accounts and then for everyone.
- 15 Oct 2025New WhatsApp Business API users begin operating under the new terms, according to respond.io.
- 18 Oct 2025TechCrunch reports the change: WhatsApp's Business API terms now bar general-purpose chatbots. It names OpenAI, Perplexity, Luzia and Poke among the providers affected and gives 15 January 2026 as the date the rule takes effect.
- 15 Jan 2026The rule applies to all existing Business API users, not only new ones (respond.io).
- 6 Mar 2026The version of the WhatsApp Business Solution Terms we read is dated 6 March 2026. It keeps the rule, with one exception: these technologies may be offered to WhatsApp users whose phone number has a European Economic Area or Brazil country code.
Meta's reasoning, as TechCrunch reports it: the Business API was built for businesses to talk to their customers, not as a distribution channel for chatbots, and general-purpose bots brought message volumes it had not planned for.
What is allowed, and what is not
Allowed: assistants for your own customers
- Customer support and FAQs
- Triage of support tickets
- Bookings, reservations and appointment reminders
- Order tracking and updates
- Notifications and delivery updates
- Lead qualification
- Authentication prompts
Use cases listed by respond.io as permitted under the 2026 policy.
Not allowed: general-purpose chatbots
- Open-domain assistants that answer anything
- AI companies using WhatsApp to distribute their assistant, such as ChatGPT or Perplexity on WhatsApp
- Any setup where the AI itself is the main thing on offer
Per respond.io, TechCrunch and the AI Providers clause of Meta's terms.
Five things that keep an assistant inside the rules
If you run an assistant on WhatsApp, or plan to, these are the five checks we use. Each one keeps it clearly on the business side of the line.
Scoped to your own services
It exists for your customers and your business: support, bookings, order status, lead qualification, notifications. Not general chat.
Why: the ban targets AI offered as the primary functionality. Your service is the point.
Answers from your own information
It replies from your prices, policies, opening hours and order data, not from whatever a model knows about the world.
Why: an assistant grounded in your information stays a business tool, and gives answers you can stand behind.
Hands over to a person
When a case needs judgement, or the customer asks, a person from your team takes over the conversation.
Why: WhatsApp's Business Messaging Policy says automation must come with “prompt, clear, and direct escalation paths”. An in-chat transfer to a person is one of them.
Politely declines off-topic requests
Asked for a poem, homework help or the news, it says what it can help with and steers back to your service.
Why: an assistant that answers anything drifts towards the open-domain chatbot the rule bans.
Keeps your WhatsApp data out of model training
If an AI provider powers your assistant, the conversations must not be used to train or improve their models.
Why: the terms let you use an AI provider as a service provider, but bar using WhatsApp data to build or train AI models, except fine-tuning one for your exclusive use.
How we build them
This is how we build WhatsApp assistants at Odysi, including the one we run for Fixlab: scoped to one business, answering from its own information, on its WhatsApp Business number through Meta's official API, and handing the conversation to a person when a case needs judgement. You can see how it works on our WhatsApp customer service assistant page.
Sources
- TechCrunch, 18 October 2025: WhatsApp changes its terms to bar general-purpose chatbots from its platform
- respond.io: WhatsApp's ban on general-purpose AI chatbots
- Meta: WhatsApp Business Solution Terms, AI Providers clause (version last modified 6 March 2026)
- Meta: WhatsApp Business Messaging Policy
This guide explains Meta's published terms as we read them on 27 September 2026. It is not legal advice, and Meta can change its terms.