WhatsApp AI Chatbots: What Meta's Latest Rules Mean for UK Businesses

Setting up an AI chatbot on WhatsApp remains permissible for your business, even after Meta restricted general-purpose AIs on its official API. What has changed is the design: the bot must exist to serve your customers regarding your business, rather than answering anything for anyone. In this guide, I'll outline this boundary, explain how to set up your customer service, and detail its costs.

TL;DR

• Section 4.7 of the WhatsApp Business Platform terms prohibits AI providers when AI is the primary functionality of the service. A company's own customer service bot is an auxiliary function and remains permitted.
• You can use ChatGPT, Claude, or Gemini behind your bot. The condition: conversation data cannot train the provider's models.
• From 1st October 2026, Meta also charges for replies within the 24-hour window: £0.0068 per message in Brazil (note: this is Meta's reference for Brazil, actual UK rates may vary slightly but this figure is used for comparison).
• The real risk of blocking comes from an open scope, lack of human handover, and a declining quality rating, not from the use of AI itself.

What Meta's Rule Prohibits, in Their Own Words

The confusion began in October 2025, when Meta announced that, from 15th January 2026, general-purpose AI assistants like ChatGPT, Perplexity, and Luzia could no longer operate within WhatsApp via the commercial API. The headline became "Meta bans AI on WhatsApp". That's not what the text says.

The text is currently found in section 4.7 of the Meta Terms for the WhatsApp Business Platform, last updated on 23rd September 2026. It states three things:

1. Who is affected: "providers and developers of artificial intelligence or machine learning technologies", as determined by Meta at its sole discretion.
2. When: when these technologies are the primary functionality, "rather than incidental or auxiliary", made available for use.
3. What remains valid: a company can retain an AI provider as its solutions provider, provided that platform data is not used to "create, develop, train, or improve" AI systems or models. Fine-tuning for the company's exclusive use is permitted.

In translation: the rule targets those who sell AI as a product within WhatsApp. A shop, school, or clinic that uses AI to serve its own customers is on the other side of the line, and Meta stated to the press, in its announcement, that this use would not be affected.

What about the regulatory challenges?

In Brazil, CADE (the Brazilian antitrust regulator) suspended the application of the new terms in January 2026 and maintained the measure in March, pending a final judgment. For businesses looking to set up an AI chatbot on their own WhatsApp, this dispute changes little: it concerns general-purpose assistants. The complete chronology is in the post about automating WhatsApp customer service with AI without being banned.

Company AI Chatbot on WhatsApp: What Remains Permitted

The question I ask every client before connecting a language model to the API is simple: if the AI chatbot on WhatsApp disappeared tomorrow, would your customer service still make sense? If the answer is yes, the AI is auxiliary. It accelerates an existing service.

The table summarises our interpretation of the terms in operation:

Use	Situation	Why
Bot that answers questions about a shop's products, delivery times, and orders	Permitted	AI is auxiliary to the company's customer service
Lead qualification from adverts and scheduling visits or appointments	Permitted	Serves the company's own customers about its own service
Support for a distance learning (EAD) course student (access, certificate, deadlines)	Permitted	Institution's service, AI as a tool
Bot using GPT, Claude, or Gemini via API with the company's key	Permitted, with condition	The provider cannot train models with conversation data
"Ask our assistant anything" as a product	Prohibited	AI becomes the primary functionality
Reselling access to an AI model via a WhatsApp number	Prohibited	This is precisely the case of an AI provider
Company bot that accepts conversations on any topic	Grey area	In practice, it becomes a general assistant and exposes the account

The last line is where most companies stumble without realising it. No one plans to create a general-purpose assistant. But a lazy prompt, such as "you are a friendly assistant from company X", allows the model to answer baking recipes and homework. Then your AI chatbot on WhatsApp starts functioning like a generic AI with a logo.

The "Auxiliary" Test: 5 Questions Before Connecting the Model

As Meta decides at its discretion what is primary or auxiliary, there is no certificate of compliance, only a defensible design. These are the questions we use to validate an AI chatbot on WhatsApp before publishing.

1. Does the bot know how to say "that's not for me"?

The system prompt needs to define the subject matter: the company's products, services, policies, and data. Outside of that, the bot politely declines and redirects to what matters.

2. Is the company speaking, not the model?

The customer converses with "XYZ Store", not with "ChatGPT". The name, tone, and signature belong to the brand. If your communication sells the model as an attraction, you push the project to the wrong side of the rule.

3. Do the answers come from your knowledge base?

The model should respond based on your content: catalogue, price list, regulations, FAQ, order status. This technique is called RAG (retrieval-augmented generation): the system retrieves the relevant snippet and provides it to the model along with the question. "Memory-based" answers are where fabrications are born.

4. Is a human just a tap away?

Human handover is the transfer of the conversation from the bot to an agent. It needs to be actionable by keyword ("agent", "human"), by detected intent (complaint, cancellation), and by failure (the bot didn't understand twice in a row).

5. Does the contract with the AI provider prohibit training with your data?

This is the explicit condition in section 4.7. Check the commercial terms of the provider's API (not those of the consumer application) to see if the data sent is used to train models, and keep this evidence. If the provider does not guarantee this in writing, change providers.

How to Set Up an AI Chatbot on WhatsApp Without Risk of Blocking

This is the sequence we follow at Agathas Web to put an AI chatbot on WhatsApp into production. It applies to a shop with one agent and an institution with ten.

Step 1: Start with the Official API, Never the QR Code

Tools that connect to WhatsApp Web via QR Code violate the application's terms and are the most common cause of banned numbers. AI plugged into them multiplies the risk. The foundation is the official API, with the account verified in Business Manager. If you are not yet familiar with the architecture, the complete guide to the WhatsApp Cloud API explains the components.

Step 2: Write the Knowledge Base Before the Prompt

Gather what a new agent would need to read on their first day: catalogue, opening hours, exchange policies, delivery times, and real questions from conversation history. Concise and updated text is worth more than an 80-page PDF. Define who updates it.

Step 3: Confine the Scope in the System Prompt

The prompt states who the bot represents, what topics it accepts, when it transfers to a human, and what it never promises (discount, out-of-policy delivery time, diagnosis). Responses in one or two messages: those who receive five consecutive bubbles tend to block the number.

Step 4: AI in Reception, Human in Queue

The AI chatbot on WhatsApp excels in reception, when the customer initiates the conversation. Proactive messages outside the 24-hour window require an approved template and opt-in, so do not let the AI send anything on its own. Also define who receives the transferred conversation and what the bot says outside of business hours: a bot that promises "I'll transfer you" and then disappears is the shortest path to a complaint.

Step 5: Test with Outsiders

Ask three external people to try and break the AI chatbot on WhatsApp: off-topic questions, discount requests, insults, audio, photos. Adjust the prompt and knowledge base until failures become polite refusals or handovers.

Step 6: Monitor Your Quality Rating Weekly

Meta gives the number a quality rating based on blocks and reports; a drop is the first sign that the bot is annoying people. If more than half of conversations end up with a human, the knowledge base is weak.

How Much It Costs to Operate an AI Chatbot on WhatsApp in 2026

The cost has three layers, and confusing them is the most common budgeting trap.

1. Message delivery by Meta. Since July 2025, charges have been per message delivered. And, according to Meta's official pricing page for non-template messages, from 1st October 2026, service messages (replies within the 24-hour window) also became chargeable, at the same rate as utility and authentication messages. In Brazil, Meta's cited reference is £0.0068 per message, without volume tiers. The 72-hour window opened by a Click to WhatsApp ad remains free of delivery charge. Details can be found in our post on official WhatsApp API conversation costs.

2. The AI model. If you use your own model (GPT, Claude, Gemini) with your key, you pay the provider for tokens, separately from Meta. If you use Meta's own agent, the Meta Business Agent, the charge is unified: £2.00 per million tokens, charged from 1st August 2026, covering AI and delivery, including within the free windows.

3. The platform. Inbox, agent queue, reports, AI orchestration. This is the tool's monthly fee.

Meta itself publishes an estimate for 10,000 AI-powered messages in Brazil:

Scenario (10,000 messages, Brazil)	Cost per message	Estimated total
Service + third-party AI, low complexity	~£0.02 to £0.03	~£268
Service + third-party AI, high complexity	~£0.09 to £0.10	~£968
Meta Business Agent, high complexity	~£0.04 to £0.05	~£400 to £500

Two messages emerge from this table. First: context size dictates the cost, so a lean knowledge base and short history reduce the model's cost. Second: compare the cost per resolved interaction, not per message. An AI chatbot on WhatsApp that resolves an issue in three messages is cheaper than a "cheap" one that needs twelve and still transfers.

Meta Business Agent, Platform with Own AI, or Custom Development

There are three ways to have an AI chatbot on WhatsApp via the official API. None is superior in all aspects.

Criterion	Meta Business Agent	Platform with Own AI (BYO)	Custom Development
Model choice	Meta's only	GPT, Claude, Gemini, DeepSeek	Any
AI charging	Per token, combined with delivery	Directly with provider, per token	Directly with provider
Inbox and human queue	Limited	Included	Needs to be built
System integration (ERP, Moodle, CRM)	Limited	Via API and webhooks	Total
Time to go live	Days	Days	Weeks or months

Meta's agent suffices for small operations and repetitive questions. The platform with its own AI serves SMEs that want to choose the model, control token costs, and have the human team on the same screen, which is the case for most clients we serve. Development is worthwhile when the bot needs to consult and record in proprietary systems, such as Moodle enrolment or a duplicate invoice.

When NOT to Implement AI on Your WhatsApp

Not all customer service should have AI, and stating this before the contract saves the client money. These are cases where an AI chatbot on WhatsApp hinders more than it helps:

• Low volume and consultative service. With 15 conversations a day, each a negotiation, an agent with quick replies resolves things better.
• Debt collection and negotiation. The wrong tone from a model in this context can lead to formal complaints. Use utility templates and human negotiation.
• Health, legal, and financial services with individual guidance. The bot can schedule and triage. It cannot provide advice.
• Non-existent knowledge base. Without reliable content, AI fabricates. First organise the information, then automate.
• No one to manage it. An AI chatbot on WhatsApp without an owner ages in weeks: prices change, and it continues to give old responses.

Common Pitfalls We Observe

• Using the consumer version of the AI provider, whose terms do not guarantee the prohibition of training.
• Budgeting only for the platform's monthly fee and forgetting Meta's delivery charges and tokens.
• Ignoring the WhatsApp API change calendar from 2026 to 2028: prices and rules change several times a year.

How Agathas Web Provides Solutions

We have developed systems and maintained servers since 2008, from critical Moodle environments to integrations with the official WhatsApp API. From this experience, Voyia, our customer service platform for the official WhatsApp API, was born. It was specifically designed for the scenario in this post: auxiliary AI, with the business in control.

What's included:

• Connection of your number to Meta's official API, with support for verification and migration from the WhatsApp Business App or another platform.
• Multi-agent inbox with queue and transfer, also available on iOS and Android apps.
• AI integration in the BYO (Bring Your Own Key) model: you connect Claude, GPT, Gemini, or DeepSeek, pay the provider directly, and Voyia orchestrates the conversation, maintains context, and transfers to a human when needed.
• Meta's rates passed on without markup, invoicing in pounds sterling, and data stored on servers in Brazil.

How it works in practice: you subscribe, we connect the number, and you configure the knowledge base, prompt, and flows using the panel's documentation and videos. If you prefer, we can set up the flow and scope prompt under a separate quotation.

How to subscribe: AI integration is included in our Professional (£349/month, up to 3 agents) and Business (£749/month, up to 5 agents) plans, with no long-term commitment. The Professional plan includes an installation fee, which our sales team will inform you of before subscription. See details and speak to our team on the Voyia page.

Conclusion: Meta's Rule is a Filter, Not a Prohibition

Meta's rule has not closed the door to AI in customer service. It has closed the door to those who use WhatsApp as a showcase for a generic assistant. An AI chatbot on WhatsApp with a scope tied to the business, its own knowledge base, human handover, and a contract that prohibits training with your data is on the right side of the line.

The next step: list the 20 most common questions that reach your WhatsApp, separate those a bot can resolve alone, and estimate the monthly volume. With this, the conversation about platform and cost becomes objective. If you'd like to do this exercise with us, learn about Voyia and speak to our team.