In October 2025 Meta changed the WhatsApp Business Solution terms to stop general-purpose AI assistants from using the business platform. The change took effect on 15 January 2026. Many businesses read the headlines as "AI chatbots are banned on WhatsApp". That is not what the terms say.
This guide explains where the line sits, how to test your own use cases against it and what to document so your AI chatbot policy holds up when Meta or a customer asks how your assistant works.
What changed on 15 January 2026
According to TechCrunch's report on the new terms, a new "AI Providers" section prohibits providers of AI technologies such as large language models and general-purpose assistants from using the WhatsApp Business Solution to offer those technologies when they are the primary, rather than incidental or ancillary, functionality. Meta decides which side of that line a service falls on.
Meta also confirmed to TechCrunch that the change does not affect businesses that use AI to serve their own customers. Its example was a travel company running a customer-service bot. That bot is still allowed.
The test: is AI the product, or does it support the product?
The practical question is simple. If you removed the AI, would there still be a business service? A store still sells, a clinic still books appointments and a logistics company still delivers. In those cases AI is incidental to the service. If the only thing a user gets from your number is open-ended conversation with a model, AI is the product, and that is the use the terms target.
| Use case | What the AI does | Likely reading |
|---|---|---|
| Store support assistant | Answers order, delivery and return questions from the store's own policies | Ancillary to retail service |
| Clinic booking bot | Collects preferred time and service, then confirms a slot | Ancillary to booking |
| Lead qualification | Asks budget and timeline, then routes to a salesperson | Ancillary to sales |
| "Ask me anything" number | Open chat with a model on any topic | AI is the primary product: prohibited |
| Paid AI-assistant subscription delivered on WhatsApp | Sells model access through the chat | AI is the primary product: prohibited |
| Homework or writing helper for the public | General answers unrelated to a business service | High risk: likely primary functionality |
This table is our reading of the published wording, not a ruling from Meta. Borderline cases should be checked against the current terms and, if needed, with your Meta business representative or solution provider.
A worked example: classifying one store's AI use
Consider a Riyadh fashion store that wants AI on WhatsApp for four jobs:
- Answer "where is my order?" using the order number and the store's shipping data.
- Explain the return policy from the store's help pages.
- Suggest sizes from the store's own size chart.
- Chat about anything the customer wants, including topics unrelated to fashion.
Jobs 1–3 are customer service and sales support for the store's own products. Job 4 turns the number into a general assistant. The safe design keeps jobs 1–3, restricts the assistant to approved store knowledge and politely declines off-topic requests with an offer to connect a team member.
Rules that applied before 2026 still apply
The AI Providers clause sits on top of the existing WhatsApp Business Messaging Policy. Your AI assistant still has to respect opt-in, template rules outside the customer service window and the expectation that customers can reach a person when automation cannot help. Our guide to WhatsApp marketing rules and opt-in covers the consent side in more detail.
- Scope: limit answers to your business, products and policies.
- Human path: give a clear way to reach staff and honour it.
- Honesty: do not let the assistant invent prices, stock or delivery dates.
- Data: collect only what the task needs and follow local privacy law, such as Saudi Arabia's PDPL.
Compliance checklist for your WhatsApp AI chatbot
| Check | Evidence to keep |
|---|---|
| Written purpose for the assistant | One paragraph naming the business tasks it supports |
| Knowledge limited to approved sources | List of documents and data the AI can read |
| Off-topic handling | Test transcript showing a polite decline and handoff offer |
| Human handoff | Routing rule, team owner and response-time target |
| Actions are controlled | List of actions (lookup, booking, tagging) and their limits |
| Review cycle | Monthly sample review of AI answers with corrections |
For handoff design, see bot-to-human handoff triggers and recovery. For building the assistant itself, the WhatsApp AI agent guide for customer service walks through knowledge, actions and testing.
What to do if you already run an AI assistant
- Read your own welcome message. If it promises to "answer anything", rewrite it around your service.
- Run ten off-topic prompts. Ask about the weather, homework and news. Each answer should decline and offer help with your service or a person.
- Check your provider's position. Ask your WhatsApp solution provider how it applies the AI Providers terms to its features.
- Record the review. Save the date, test prompts and changes you made.
How WA Smart Sender fits
WA Smart Sender runs on the official WhatsApp Cloud API. Its WhatsApp automation and AI agent features are built for business tasks such as support, booking, order updates and lead routing, with handoff to a shared team inbox. You still decide what your assistant knows and how it answers, so the checklist above applies. Plans are on the pricing page.
FAQ
Are AI chatbots banned on WhatsApp in 2026?
No. The terms bar general-purpose AI assistants where AI is the primary service. Businesses can still use AI that supports their own customer service and sales.
Can I connect ChatGPT or another model to my WhatsApp Business number?
Using a model behind your support or booking flow is different from offering open model access to the public. Keep the assistant limited to your business tasks and keep a human path.
Who decides if my use is allowed?
Meta. The terms say the primary-versus-incidental question is determined by Meta in its sole discretion, so document your purpose and limits.
