A WhatsApp chatbot is an automated assistant that answers customers inside WhatsApp at any hour, connected to the company's own systems: the CRM, the order management tool, the booking calendar. The customer writes the way they would text a friend, and the bot understands the request, pulls the data it needs and replies, or hands the chat to an agent when the case calls for it. For a business, it moves part of the work that now goes through phone, email and web forms onto a channel people open dozens of times a day.
2026 changed a lot for anyone running a bot on WhatsApp. Meta launched its own Meta Business Agent, removed general-purpose AI assistants such as ChatGPT from the platform, and started charging for customer service messages above a free allowance. If you are evaluating a WhatsApp Business chatbot today, you need to understand these rules before you pick the technology.
This guide covers how a WhatsApp chatbot works, what it can do for sales, support and marketing, what it costs under Meta's current pricing, and how to choose between Meta's agent and a dedicated platform.
Table of contents
- How a WhatsApp chatbot works
- WhatsApp Business App or Business Platform
- What a WhatsApp chatbot can do for your business
- Meta's rules for WhatsApp chatbots in 2026
- Meta Business Agent or a dedicated platform
- How to integrate a WhatsApp chatbot in six steps
- Where to start
- WhatsApp chatbot FAQ
How a WhatsApp chatbot works
A modern WhatsApp chatbot is an AI agent: software that uses a large language model (LLM) to understand what the customer is asking, even when the message is badly typed, full of abbreviations or mixes two requests at once. The model on its own knows nothing about your company, though. So the agent works from a knowledge base (catalogues, price lists, terms of sale, internal FAQs) and connects through APIs to the systems that hold the real data: an order's status, a room's availability, a bill's balance. This technique is called RAG, Retrieval-Augmented Generation, and it makes the bot answer with current information instead of made-up sentences.
The connection to WhatsApp goes through the WhatsApp Business Platform, Meta's official API. The chatbot platform receives each customer message, processes it and sends the reply through the API, either directly or via a Business Solution Provider (BSP) such as Twilio. The bot can read text, voice notes and images, and reply with text, buttons, list menus, documents and payment links.
Here is a concrete example. A customer of an energy company writes at 10:40 pm: "March bill is way too high, did you change my meter?". The chatbot spots two requests, checks the customer ID, reads the amount and the last meter reading from the billing system, explains that the bill includes an adjustment and attaches the PDF. If the customer pushes back or asks to pay in instalments, the bot opens a ticket and hands the chat to a human agent the next morning, with the full history already written. Sorgenia uses this pattern for both pre-sales and post-sales, on the web and on WhatsApp.
WhatsApp Business App or Business Platform
There are three versions of WhatsApp, and only one lets you connect an AI chatbot with real integrations:
- WhatsApp Messenger, the app everyone uses to text friends and family;
- WhatsApp Business App, free, built for shops and freelancers who reply by hand from a smartphone, with automatic greeting and away messages;
- WhatsApp Business Platform (the Business API), for mid-sized and large companies: it runs through an API, handles thousands of conversations in parallel, connects to CRMs and back-office systems, and sends proactive messages with Meta-approved templates.
If the goal is to automate support or sales, choose the Business Platform. The Business App works as long as a person is typing the answers, but it cannot handle high volumes and does not integrate with company systems.
What a WhatsApp chatbot can do for your business

Use cases fall into two groups. In the first, the customer writes and the bot answers (support, bookings, sales). In the second, the company starts the conversation with a template message (marketing, reminders, notifications). The two groups follow different rules and prices, as the section on Meta's rules explains.
Customer care and support
This is the most common use case. The chatbot answers frequent questions, gives the status of an order or shipment, sends documents and solves recurring problems. Requests it cannot handle go to an agent along with a summary of the conversation. Centrale del Latte di Torino runs an AI agent for customer service on WhatsApp and on its website. To see how this changes a support team, read our guide on AI agents for customer care.
Bookings and reminders
The customer books, moves or cancels an appointment in the chat, without calling and without downloading an app. The day before, they get a reminder with "confirm" and "reschedule" buttons, which cuts no-shows. It works for clinics, garages, salons, hotels and tour operators: Hello Vacanze uses it for travel booking. We wrote a full guide on the appointment booking chatbot.
Selling in the chat
The bot recommends a product after two or three questions, shows cards from the catalogue, handles doubts about sizes, delivery times and returns, and sends a payment link. For online retailers it is the most direct way to bring the cart into the conversation, and one of the most practical applications of AI for ecommerce. In B2B the same pattern collects repeat orders: MW.FEP takes customer orders on WhatsApp instead of by email and phone.
Outbound marketing and proactive campaigns
Here the company writes first, with a Meta-approved template and only to people who opted in: a product launch, a reserved offer, an event invitation, a feedback request after a purchase. The advantage is that the customer's reply opens a conversation the chatbot handles right away, instead of landing in an inbox nobody reads. TLC Marketing used a WhatsApp chatbot to engage the partners of its loyalty campaigns and raised the redemption rate by 19%; the details are in the TLC case study.
Engagement and gamification
Quizzes, prize draws, timed challenges, countdowns before a sale. These formats work well on WhatsApp because the message arrives in the same app as personal chats and replying takes one tap. Use them sparingly: every marketing message has a cost, and people who feel spammed block the number.
Meta's rules for WhatsApp chatbots in 2026
In 2026 Meta changed both the terms of use of the Business Platform and the way it bills for it. Three changes affect anyone designing a WhatsApp chatbot.
General-purpose AI assistants are out
Since January 15, 2026, the WhatsApp Business Solution terms bar AI providers when the AI assistant is the main function of the service. In practice, all-purpose chatbots such as ChatGPT, Copilot or Perplexity can no longer be distributed on WhatsApp. Meta told TechCrunch that the ban does not affect businesses using AI to serve their own customers: a support, sales or booking chatbot built for a company is still fully allowed.
Italy is a special case. After interim measures by the Italian Competition Authority (AGCM), Meta suspended the ban for Italian users and, from February 16, 2026, charges AI assistant providers for every generated response, as reported by ANSA. A company running a chatbot for its own customers does not pay this fee; the standard per-message rates below apply.
How much a WhatsApp chatbot costs
The cost of a WhatsApp chatbot has two parts: the platform that runs it (licence, design, integrations) and Meta's fees for messages. Meta's rates are public and change several times a year. According to Meta's official pricing documentation, they work like this:
- Meta charges per delivered message, at a price set by the message category and the recipient's country code. Messages customers send to the business are free;
- template messages, which the business can send at any time, come in three categories: marketing (the most expensive), utility (updates on orders, appointments, accounts) and authentication (verification codes);
- free-form replies inside the 24-hour window after the customer's last message are "service" messages. From October 1, 2026, they are charged beyond the first 1,000 per month per phone number, and utility templates sent in reply to the customer are charged from that date too;
- when the conversation starts from a Click-to-WhatsApp ad, messages in the linked free entry point window are not charged;
- messages generated by Meta Business Agent have their own usage-based rate: $2 per million tokens.
Rates also differ by market: in Italy, for example, the marketing rate went up on July 1, 2026. Our practical advice is to design the chatbot so it resolves each request in a few messages, use marketing templates only where they bring a measurable return, and check Meta's CSV rate card before budgeting a campaign.
Meta Business Agent or a dedicated platform
On June 3, 2026, Meta announced Meta Business Agent, Meta's AI agent for businesses on WhatsApp, Messenger and Instagram, available globally. Meta says more than one million businesses already use it, and that there are more than one billion active threads with businesses on its apps every day. The agent answers questions about the business, recommends products from the catalogue, books appointments, qualifies leads and connects to systems such as Shopify and Zendesk. Getting started is free, with paid subscriptions coming in the following months. Larger companies get the Meta Business Agent Platform, which works alongside the Business Platform.
For many small businesses it is a good starting point: you can set it up in minutes from your business profile and catalogue. The limits show when processes get more complex. The agent runs on Meta's infrastructure and under Meta's rules, it is tied to Meta's apps, and customisation stops where the product stops.
A dedicated AI agent building platform such as Crafter.ai makes sense when:
- the chatbot must work across several channels with the same logic: WhatsApp, website, app, Microsoft Teams, phone;
- you need deep integrations with internal systems (ERP, CRM, industry software), not just ready-made connectors;
- you operate in a regulated industry and want control over where data is processed: Crafter.ai complies with GDPR and the AI Act, with data residency on European servers;
- the conversation flow needs precise rules, such as identity checks, approvals, routing to different departments and a brand voice tuned sentence by sentence;
- you want to switch language model or vendor without rebuilding everything.
Our recommendation: if you run a shop with a catalogue and repetitive questions, try Meta Business Agent. If WhatsApp is one channel of your customer service, or if the bot has to read from and write to your systems, choose a platform that connects to WhatsApp through the API but keeps the logic and the data under your control. The Crafter.ai WhatsApp chatbot is built for this second scenario. We also covered it in the webinar Omnichannel AI Agents on WhatsApp.
How to integrate a WhatsApp chatbot in six steps
- Define the scope. Pick the five to ten requests that weigh most on phone and email today and start there. A bot that does three things well beats one that does twenty badly.
- Activate the WhatsApp Business Platform. You need a verified Meta Business account and a phone number. You can use a new one or connect the number you already use with the WhatsApp Business app: Meta allows this through coexistence, with some limits such as group chats. The chatbot platform or the BSP walks you through registration.
- Build the knowledge base. Gather FAQs, terms, price lists and procedures, and decide which systems the bot must reach through APIs.
- Design the conversation and the handover. Decide when the bot should stop and pass the chat to a person, and with what information.
- Prepare the templates. If you plan to send reminders or marketing messages, write the templates and submit them to Meta for approval, choosing the right category because it sets the price.
- Test, launch, measure. Try the bot with colleagues and a small group of customers, then track resolution rate, handovers, cost per conversation and satisfaction every week.
If you want to see how an agent is built without writing code, our guide on how to create a chatbot starts from scratch.
Where to start
The most common mistake is starting with outbound marketing. It is the most visible part, but also the most expensive per message and the riskiest for the number's reputation. Start with support instead: customers already write to you, the bot takes load off the team from the first week, and the conversations it collects tell you which proactive messages would actually make sense. Before signing with a vendor, ask three things: how it handles the handover to agents, where conversation data is stored, and how it shows you Meta's costs message by message. To see how it works on your own case, the WhatsApp chatbot for business page lets you book a demo or try the platform for free.
WhatsApp chatbot FAQ
What is a WhatsApp chatbot?
It is an automated assistant that answers your customers on WhatsApp for you, day and night. It uses artificial intelligence to understand questions and connects to your business systems to give precise answers, like an order's status or available time slots.
Can I use a chatbot with the free WhatsApp Business app?
Not with real integrations. The free app only has automatic greeting and away messages, plus Meta Business Agent. To connect a chatbot to your systems you need the WhatsApp Business Platform, that is, the API.
How much does a WhatsApp chatbot cost?
It depends on two things: the platform you use and the messages. Meta charges for each delivered message based on its category (marketing, utility, authentication, service) and the recipient's country. From October 1, 2026, service messages are free up to 1,000 per month per number, then they are charged. Messages your customers send you cost nothing.
Are AI chatbots banned on WhatsApp after January 15, 2026?
No. General-purpose AI assistants like ChatGPT, offered as a standalone service, are banned. Chatbots a business uses for support, sales or bookings with its own customers are still allowed.
What is Meta Business Agent?
It is Meta's AI agent for businesses, launched in June 2026 on WhatsApp, Messenger and Instagram. It answers customers, recommends products and books appointments. It is free to start and will move to a subscription.
Is a WhatsApp chatbot GDPR compliant?
It can be, and it depends on the platform. You need a legal basis for the messages, consent for marketing, a notice telling customers they are talking to an AI system, and a vendor that processes data in Europe. With Crafter.ai, data stays on European servers.
Does the chatbot replace human agents?
No. It handles repetitive requests and passes complex or sensitive cases to a person, with the conversation already summarised.
How long does it take to go live?
Meta Business Agent takes a few minutes to set up. A chatbot integrated with your systems usually takes a few weeks, spent mostly on preparing content and integrations, not on writing code.





