How to connect ChatGPT to WhatsApp Business

What Meta allows

Since 15 January 2026, section 4.7 of Meta's WhatsApp Business terms says providers of “large language models, generative artificial intelligence platforms, general-purpose artificial intelligence assistants, or similar technologies” are “strictly prohibited” from using the WhatsApp Business Platform to offer those technologies “when such technologies are the primary (rather than incidental or ancillary) functionality being made available for use, as determined by Meta in its sole discretion”. The exception is countries where Meta is legally required to allow it. The terms were last modified on 23 September 2026, and Meta's AI providers page gives 15 January 2026 as the date the rule took effect.

That rule is aimed at assistants like a public ChatGPT contact in your chat list. Your own business answering its own customers is a different use: TechCrunch reported that Meta said the change doesn't affect businesses that are using AI to serve customers on WhatsApp. The same terms say you may keep an AI provider as your solution provider, as long as WhatsApp Business Platform data isn't used to create, train or improve AI models, apart from fine-tuning one for your exclusive use. Because Meta decides case by case, keep the AI in the role of your business's front desk, not a general assistant.

The three routes

1. Build the connection yourself

You need a Meta business account and an app, a phone number on the WhatsApp Business Platform, and a server of your own. Meta sends each customer message to that server through a webhook. Your server passes it to an AI model's API and sends the answer back through the Cloud API. It works, and it's developer work.

• Effort: developer work. A basic reply loop is the quick part; remembering each customer, handing over to a person and a team inbox take much longer.
• You maintain: the server, Meta app permissions (and App Review, if it applies), template approvals, error handling, and the AI prompts.
• Cost drivers: developer time, hosting, the AI provider's per-use bill, and Meta's message fees.

2. A no-code automation tool

Generic automation tools can connect a WhatsApp number to an AI model with drag-and-drop steps. You still need a number on the Platform, and you build the logic yourself, one step at a time.

• Effort: quicker to start than building it yourself; every new case is another step you build.
• You maintain: the workflow, and every edge case it doesn't cover.
• Cost drivers: the tool's subscription, the AI provider's usage bill, and Meta's message fees.

3. A WhatsApp AI agent made for business chats

A product built for this connects the number and handles the 24-hour window and templates. You describe your business and review what it does.

• Effort: connecting a number and writing down what your business sells and what you need to know from a customer.
• You maintain: your own business information.
• Cost drivers: the provider's fee and Meta's message fees.

What it takes to run any of them

• The 24-hour window. Meta calls it a customer service window: a 24-hour timer that starts when a customer messages you. Inside it you can reply freely.
• Approved templates. To start a conversation, or write after the window closes, you send a template Meta has approved.
• Message fees. Whichever route you pick, Meta bills you for delivered messages from your business. The pricing calculator estimates the bill by country, and the October 2026 pricing guide covers what changed.

What a bare chatbot lacks

A model wired straight to WhatsApp answers the message in front of it. For a business, the gaps show quickly.

• It doesn't remember each customer across chats unless you build that.
• It asks whatever the model thinks of, not your qualifying questions, one at a time.
• It has no handover to a person with a reason attached.
• There's no shared inbox where your team can take over and hand back.
• There are no lead notes or scores.
• Unless you write the rules into its prompt, nothing tells it not to invent a price or a discount.

How Zymbler does it

Zymbler is route three: a WhatsApp AI agent for business chats, not ChatGPT, built on the official WhatsApp Business Platform. You describe your business and write your questions in plain words. Here is how the agent for Horizon Holidays, a made-up travel agency, handles a chat.

Next time Maya writes, the agent already has those notes. Your team sees them in a shared inbox and can take over a chat and hand it back. Each lead gets a score from 0 to 100, and you get an email when a lead is ready. The agent states only the prices and policies you've written down and is told never to invent a discount. It can reply in each customer's own language; you choose this when you set up the agent. Where Meta allows it, it works alongside the WhatsApp Business app on the same number (see WhatsApp coexistence).