WhatsApp Chatbot with AI: The Complete Guide for 2026

WhatsApp has over 3 billion monthly active users and, in most of Europe, Latin America, India, and increasingly the US, it has become the default channel customers reach out to before email. Setting up a WhatsApp chatbot is no longer an experiment: it is the most direct way to stop losing sales at 11pm or on a Saturday afternoon, on the channel your customer already uses every day.

But the topic is full of confusion. WhatsApp Business this, WhatsApp Business API that, templates, the «24-hour window», and the famous «free chatbot» that turns out not to be free. This guide is the no-hype version: what a WhatsApp chatbot really is, what AI can actually do today, how much it costs, and how to set one up without wasting two weeks. We have spent months building our own WhatsApp chatbot at Bravos AI, we are a verified Meta Tech Provider, and our customers can connect their WhatsApp number in minutes, not days.

60-second summary

If you're in a hurry, here's what you need to know. The rest of the article goes deep on each point with honesty:

• A WhatsApp chatbot is a program that replies to your customers on your WhatsApp Business number, 24/7, without anyone having to be there. If it has AI behind it, it doesn't rely on rigid buttons: it understands the question in plain language and answers using your business information.
• WhatsApp Business (the mobile app) and WhatsApp Business API (the technical connection used by chatbots) are two different things. The app is free and operated by a person; the API is for connecting software like a chatbot and has its own per-message cost.
• Since July 2025, Meta no longer charges for conversations started by the customer. If your chatbot only replies (no outbound marketing), Meta's cost is basically zero. You only pay the chatbot platform.
• The only thing that's truly free is the fixed away message of the official WhatsApp Business app. Anything with real conversational AI costs money. The question is how much.
• A WhatsApp chatbot with real AI costs between $23 and $200/month depending on platform and volume, without counting outbound marketing messages.
• Setup takes from minutes to several days depending on the platform. If the platform is a verified Meta Tech Provider (we are), it's minutes. If they work through a BSP intermediary, it can be days. If you build it yourself with n8n or Make, plan for 13–26 hours the first time.

What is a WhatsApp chatbot

A WhatsApp chatbot is a program that receives the messages customers send to your business WhatsApp number and replies on its own, with no person typing on the other side. If it's built with modern AI, it is not a button menu: it understands what the customer wants even with typos, in another language or in very informal wording, and replies with real information from your business.

In 2026, a typical WhatsApp chatbot does three things:

• Answer frequently asked questions: hours, prices, location, payment methods, return policy, allergens, service conditions.
• Find specific information about your business: «what treatments do you offer for back pain?», «what does your full coverage insurance include?», «do you ship to Hawaii?».
• Capture leads and collect data so your team can take over: when someone shows interest (or when the machine can't close the case), the bot asks for email or phone and saves that contact in your dashboard.

What a serious WhatsApp chatbot doesn't do:

• It doesn't start conversations on its own to sell you things (that's outbound marketing and it has strict rules — we'll cover it below).
• It doesn't access your personal WhatsApp conversations: it only works with the WhatsApp Business number you connect.
• It doesn't replace the human team on complex cases. What changes is that your team stops answering the trivial stuff.

WhatsApp Business app vs WhatsApp Business API: the confusion everyone has

This is the most confused topic in the WhatsApp chatbot world. You google «WhatsApp Business chatbot» and you find two things with almost identical names that have nothing to do with each other. Let's break it down.

WhatsApp Business app: the mobile app

It's the app you download on your phone, different from regular WhatsApp, designed so a sole trader or small business can have a separate number from their personal one. It includes a business profile, a basic catalogue, customer labels and quick replies. It's free and operated by a person at the business.

What looks like a chatbot inside the app is actually two very limited things:

• Greeting message: fixed text when someone messages you for the first time.
• Away message: fixed text when someone writes outside your business hours.

That's not AI, that's not a conversational chatbot, it doesn't understand what the customer says. It's an automated sign. Enough for many sole traders; not enough for a small business with two people in customer service. If you want to understand the different levels of automated reply, we cover them in WhatsApp auto reply: 3 real options.

WhatsApp Business API: the chatbot connection

It's the technical channel Meta offers so external software (an AI chatbot, a customer service dashboard, a CRM) can send and receive messages on your WhatsApp number. It has no app, you don't see it on a phone: it lives inside the dashboard of whatever platform you're using.

The original «Business API» (the old «On-Premise API», hosted on your own server) is barely used anymore. What everyone today calls «Business API» is actually the WhatsApp Cloud API, the version hosted by Meta and launched in 2022. The difference between «Business API» and «Cloud API» is only where the infrastructure runs; for the end customer they are the same thing.

At a glance

Bottom line: if you just want an automated sign about your hours and you'll reply by hand, the app is enough. If you want a chatbot that understands questions and replies with real data from your business without anyone typing, you need to connect your number to the API through a platform.

What an AI chatbot can do on WhatsApp (and what it can't)

This is where we have to be honest. Ads promise the world; the reality of a chatbot with real AI on WhatsApp in 2026 is better than people think for some things and worse for others.

What a decent AI chatbot does well

• Answer questions with your business information. If you upload your website, PDFs or a spreadsheet, it finds the answer there and replies in plain language. Quality varies a lot by platform: the good ones don't make things up when they don't have the data; the bad ones do.
• Remember what was said within the same conversation. If you ask the price of product X and then say «and in size L?», it knows which product you mean. This is the bare minimum a modern conversational AI should offer.
• Be available 24/7. No hours, no holidays, no waiting. The difference between replying in 30 seconds and replying the next day is the difference between closing the sale and not closing it.
• Collect customer data when needed. If the question needs human attention or the customer asks for a person, the bot asks for phone or email and drops the contact in your dashboard so your team can pick it up later.
• Honestly say when it doesn't know (on serious platforms). If you haven't uploaded the return policy, it won't make one up: it says it doesn't have it. Platforms that invent answers are a real problem in the market — worth testing before signing up.

What it doesn't do (or rarely does)

• It doesn't pick up the phone. It doesn't replace a call for an angry customer, a complex negotiation or an emotionally delicate case.
• It doesn't book appointments in your calendar by itself on most platforms. What it can do is share available slots, send a link to your booking system, or take the data so your team confirms the appointment. The «books your full calendar» integration exists on very few platforms and usually requires extra configuration.
• It doesn't close complex sales by itself. For a $50 sneaker purchase, sure, it can walk the customer to checkout. For an $8,000 kitchen renovation, no. It filters interest and drops the data in your dashboard so your team can close it.
• It doesn't send messages to your customers on its own unless you configure it to. And when outbound is configured, WhatsApp imposes the «24-hour window» (we cover it below), which limits that type of message.
• It doesn't process orders or payments by itself. It can share your store's checkout link; it doesn't act as a payment gateway.

If the platform you're looking at promises things their product doesn't do (like a fully autonomous agent that closes sales, books calendar appointments and processes payments without intervention), be skeptical and ask for a demo with your case. The technical reality of most WhatsApp chatbots in 2026 is much closer to a great receptionist than to an autonomous salesperson.

Are there free WhatsApp chatbots? What's actually free and what isn't

A lot of search volume around WhatsApp chatbots adds «free» somewhere in the query. The short, honest answer: a conversational WhatsApp chatbot with real AI that's free forever doesn't exist. What does exist is several things marketed as free that sometimes are and sometimes aren't. Let's separate them.

What's actually 100% free

• The away message and greeting of the WhatsApp Business app. Fixed text, no AI, doesn't understand the customer. The most basic thing and, for many sole traders, the only thing they need.
• Trying several platforms for a limited time. Many offer free trials of 7 to 14 days, including ours. That lets you see if it fits before paying. It's not a «permanent free plan».

What's marketed as free but has small print

• «Free» plans of web chat platforms (Tidio, Crisp, Freshchat, etc.). The free plan covers their web widget chat, not WhatsApp: to connect WhatsApp Business API you need a paid plan. And even when the free plan promises AI, that AI is usually quite limited — we checked it by asking their own official chatbots in this experiment.
• Twilio «sandbox» trials. Yes, you can connect WhatsApp for free for testing, but only with numbers you have registered and with pre-defined templates. Not viable for a real business.
• Building it yourself with self-hosted n8n. Free n8n (hosted on your own server) does save you the subscription, but it costs you the server ($5–20/month), the OpenAI or Claude model you use ($5–50/month depending on volume) and the hours you put in. We break it down in this comparison with n8n, Make and Zapier.

What Meta always charges for

This is important and almost no one explains it: Meta charges for some WhatsApp Business API conversations, regardless of the platform you use. That's a Meta cost, not a platform cost. Since July 2025 the model changed:

• Conversations started by the customer (service conversations): free and unlimited. If the customer messages you first, neither Meta nor anyone else charges for the message.
• Conversations started by your business (marketing, reminders, notifications): these have a cost. In the US, marketing messages run around $0.025 per message; in other regions it varies. If you only use the chatbot to reply to what comes in, this cost doesn't affect you.

Honest bottom line: if you want a real AI chatbot on your WhatsApp Business, you're going to pay between $23 and $200/month to the platform. We sit in that range: from $23/month on the Starter plan. That's what it costs to have a maintained system, officially connected to Meta, with a decent AI model behind it. «Free forever» with real AI doesn't exist.

How it connects: signup, templates, and the 24-hour window

The signup process for a WhatsApp chatbot has parts that almost no one bothers to explain, but knowing them saves you headaches. These are the concepts you'll hear sooner or later.

Do I need to verify my business with Meta?

For a chatbot that only replies to inbound, no. Up until 2024 you did have to pass business verification in Meta Business Manager before using the API. That's no longer the case: you can connect your number, receive messages and reply without passing any verification. Replies to customers who message first are unlimited in this mode.

Verification is only required if you're going to use WhatsApp for heavy outbound marketing: without verification, Meta lets you send business-initiated templates to up to 250 unique customers in a rolling 24 hours. If you complete verification (or if your message quality is high enough that Meta promotes you automatically), that limit goes up to 100,000 per day.

Because we are a verified Meta Tech Provider, connecting your number through the Bravos AI platform takes minutes thanks to Embedded Signup. On platforms that work through intermediaries (so-called BSPs, «Business Solution Providers»), signup can take longer.

Message templates

If your business wants to start a conversation with a customer (sending an appointment reminder, a shipping update, an offer), Meta requires that message to be sent with a pre-approved text, called a template. You create the template and submit it to Meta for review; they approve or reject it within hours. This is what prevents businesses from using WhatsApp as spam.

Replies to a customer who messaged you first do not need a template. Your chatbot can reply with whatever it wants, in free language, with no approval. This is what makes a reactive chatbot (like ours) much simpler to put live than one that pushes outbound marketing messages.

The 24-hour window

When a customer messages you, Meta opens what it calls a 24-hour service window. During those 24 hours your business (or your chatbot) can reply freely, with no template and no cost. Once 24 hours pass, if you want to message that customer again, you can only do it with an approved template. So for regular customer service, the 24-hour window is not an issue: the bot replies in seconds, well before it closes.

Phone number quality with Meta

Meta tracks the behavior of your number. If customers block you or mark your messages as spam, your «quality» rating drops and Meta can limit how many templates you can send per day. This only affects you if your platform allows outbound and you're using it: in that case, segment well before sending. If your chatbot only replies to what comes in (like ours), the quality problem doesn't apply to you, because replies to customers don't count toward that score.

How to build a WhatsApp chatbot step by step

If you have the basics down and you want to set it up, here's the short path. The long version, with the Meta panel walkthrough, is in our dedicated guide how to build a WhatsApp AI chatbot in 2026 without code.

1. Decide what you want the bot to do. Answer FAQs, search your catalogue, collect data for an appointment or for your team to take over. Keeping a tight list prevents you from drowning in features you don't need in month one.
2. Gather your business information. Website text, PDFs of your menu or catalogue, a spreadsheet with products or services, return policy. If you have an online store on Shopify or PrestaShop (or your own system with a webhook), some platforms (ours included) connect directly and the catalogue syncs automatically, without you having to maintain a separate file. A chatbot is only as good as the information you give it, as we detail in this guide on keeping your chatbot up to date.
3. Sign up for a platform. There are dozens; we compare the 13 most relevant in this comparison of the best WhatsApp chatbots. The choice depends on your budget, your language needs and whether you'll do outbound marketing or just inbound support.
4. Create the bot and train it on your information. On modern platforms this means uploading files or pointing it to your website URL. The platform processes it and can start replying based on that.
5. Connect your WhatsApp number. The platform walks you through the Meta Business signup and number assignment. If the platform is a verified Meta Tech Provider (like us), the connection takes minutes. If not, it can take longer. Business verification is only required if your platform allows outbound and you later want to send templates to more than 250 contacts per day.
6. Test with real conversations before exposing it to customers. Throw weird questions at it, typos, words in other languages, scenarios where you know information is missing. Adjust as needed.
7. Activate and observe the first week. Watch what people actually ask, what the bot gets wrong, what edge cases slip through. Most platforms (ours included) let you see the full conversation history so you can refine the loaded information.

How much a WhatsApp chatbot really costs

Price depends on three things: the platform you choose, the messages you send, and, if you build it yourself, the AI model behind it. Let's break them apart.

Platform cost

Platforms with a real AI WhatsApp chatbot cover different cases. These are the typical ranges in June 2026, with examples of platforms that fit each one:

The detail for each one, with prices verified directly on their official pages, is in the comparison of the 13 most relevant platforms.

Meta message cost

Since July 2025, conversations started by the customer are free forever. You only pay Meta when your business starts the conversation with a template. For reactive customer service, this cost is basically zero. If you run mass marketing campaigns on WhatsApp, budget around $0.025 per marketing message in the US (it varies by region and category).

The hidden markup

Some platforms charge a markup on top of Meta's cost for every message. Wati, Interakt and Twilio apply margins from 2.8% to 40% depending on plan and volume. Others don't apply a markup and pass Meta's price through. Worth asking before you sign; we break it down platform by platform in the comparison.

For a broader pricing comparison, not just WhatsApp, see our complete guide on how much an AI chatbot costs in 2026.

AI chatbot vs button flows: why the difference matters

When you search for «WhatsApp AI chatbot» or «WhatsApp ChatGPT chatbot», what you're asking for is a chatbot that understands what the customer says in free language. That's not the same as a button menu that many platforms still sell as a chatbot.

Button flows (the classic chatbot)

It works like this: when the customer writes, the bot offers options («1) Hours, 2) Prices, 3) Book»). The customer picks one, the bot gives a fixed answer, offers more options, and so on until they leave or the flow ends. If the customer writes anything outside the menu, the bot doesn't understand and usually replies with a generic «I didn't catch that».

This is what ManyChat, AiSensy on its base plan or Landbot do. It's not bad; for highly structured processes (surveys, polls, first-line filtering) it can work. But most of your customers don't want to navigate a menu: they want to ask what they need in their own words.

Conversational AI chatbot (what people mean today by «AI chatbot»)

The customer writes whatever they want, the way they would write to a person, and the bot understands. If they ask «what time do you open on Saturday?», the bot finds that information in what you uploaded and replies. If they ask about your return policy, the bot pulls it from your policy document and explains it in its own words, not as a copy-paste of the PDF.

Under the hood, what the bot does is combine your information with a large language model (popularly called «the engine of the chatbot»): GPT-4, Claude, Gemini or another. The platform's job is to make sure the model doesn't make things up about your business: it only passes your information as reference.

This is what we do at Bravos AI, what Tidio does with Lyro, what Wati does with Astra, what Intercom does with Fin. The difference between them is which model they use, how well they manage your information, and what they charge. If you want to go deeper, our rule-based vs AI chatbot comparison covers it.

Honest recommendation: in 2026, launching a brand-new chatbot based purely on buttons rarely makes sense. Customers are used to talking with ChatGPT and Google's AI. If your chatbot doesn't understand them when they ask normally, they'll abandon it by the second interaction.

Use cases by sector: two examples

To show you what this actually looks like in production, here are two example conversations of the kind that hit businesses with a connected bot.

Retail: outdoor gear store

The customer wants boots for a specific trip. Watch what happens when the bot knows the catalogue and the product details:

What this bot does well: filters the catalogue by size, by use case (the trip drives the recommendation), compares two specific models by a subjective feature (wide vs narrow last), and closes by sending the direct product link. It doesn't push the sale, doesn't ask for personal data if it doesn't need to.

Professional services: dental clinic

The patient has a specific question about a treatment. Watch how the bot informs with clinic data and sends them to the booking system without booking on her behalf:

What this bot does well: replies with clinic data (prices, sessions, warranties), doesn't get into recommending medical treatments, doesn't decide for the patient and, most importantly, sends her to the clinic's booking system instead of promising a reservation the machine can't confirm. If the patient would rather be called, the bot takes the phone number for reception to ring back.

Other sectors that fit a WhatsApp AI chatbot well: restaurants (menu, allergens, reservations), physical retail (hours, stock, location), real estate agencies (property filters), businesses with wide catalogues (filtering and product detail), training academies (prices, dates, terms), online stores (returns, order tracking, recommendations).

Common mistakes when launching a WhatsApp chatbot

We've spent months watching small businesses, sole traders and agencies launch WhatsApp chatbots. These are the most common pitfalls and how to avoid them.

1. Uploading incomplete information and blaming the bot

If your website doesn't list the return policy, the bot isn't going to make one up (that's a good thing). But if you launch the bot without that content, you'll have customers asking things the bot can't answer. Before going live, double-check you have everything a real customer might ask. If something's missing, write it, upload it and verify.

2. Trying to reuse the same number you already have in the WhatsApp Business app

A WhatsApp number can only be in one place at a time: either in the WhatsApp Business app on your phone, or on the API connected to a chatbot. Not both. If you want to move the number you use in the app over to the chatbot, you have to migrate it (Meta walks you through it) and that wipes the history you had in the app. If you don't want to lose that history, the easiest path is to get a new dedicated number for the bot.

3. Treating WhatsApp like email (when your platform allows it)

Only applies if your platform allows outbound template sending, not if your chatbot only replies to inbound. On those platforms, sending marketing templates to 5,000 contacts who didn't opt in ends badly: customers block, Meta drops your quality rating, and your sending limits start to shrink. If you're going to do WhatsApp marketing, segment well before sending and start with fewer contacts.

4. Not reviewing conversations in the first week

The first 50 conversations of your bot are gold: that's where you see the real questions people are asking and the gaps in your information. If you don't look at them, you can't improve the bot. Most platforms, ours included, let you see the full conversation history and filter the ones the bot got wrong.

5. Expecting the chatbot to fully replace the human team

A well-trained chatbot handles most of the routine queries on its own. The rest still need a person. If you frame it as «I'm replacing my customer service team», you're in for a disappointment. If you frame it as «my team stops answering hours, prices and availability and spends their time on what matters», it fits perfectly.

6. If you're going to do outbound, not checking the per-message markup

If you're only doing inbound, this one doesn't apply to you. But if your platform allows outbound and you're going to send templates, watch out: a platform at $23/month can end up more expensive than one at $60/month if the first one charges a 25% markup on every Meta template and the second doesn't. Always look at both numbers: monthly fee and per-message pricing model.

When it's worth building it yourself with n8n, Make or Zapier

If you're technical (or work with someone who is), there are several routes: automation tools like n8n, Make or Zapier connected to OpenAI or Claude's API, or a more developer-oriented bot builder like Google's Dialogflow. We did this ourselves as an exercise with n8n and we wrote it up honestly in this comparison with n8n, Make and Zapier.

The summary: it makes sense if you're already heavy users of that tool for other things, you have a technical team maintaining flows in production, and you need something very custom that doesn't fit any closed platform.

It doesn't make sense if you're a sole trader or a small business without technical people, if you don't have anyone who can put in the 15-25 hours the first build takes (plus the maintenance after: images, audio, conversation memory, 60-day token expiry), or if you want it live this month and not next.

How we do it at Bravos AI

If you made it this far, we owe you a quick run-down of what we do. At Bravos AI we build AI chatbots for small businesses and larger companies on a custom plan that want to automate customer support over WhatsApp and on their website without going through expensive agencies or building anything technical. Here's what sets us apart, told without the marketing fluff:

• Verified Meta Tech Provider. That means connecting your WhatsApp with us takes minutes via Embedded Signup, no BSP intermediary, no 60-day token expiry. It's the official Meta model.
• Reactive bot, not proactive. We don't start conversations on your behalf, we don't send campaigns to your base, we don't push you toward outbound marketing. The bot replies to whoever messages you. Safest way to avoid burning your number with Meta.
• Native integrations: Shopify, PrestaShop, Resales Online (real estate) and custom webhook. We hook into your catalogue or your system directly and the bot stays current without Zapier in the middle. If your business lives in another system, the webhook brings it in without custom development.
• Exact search on your catalogue, not approximate. If your customer asks for size 8 under $200, the bot finds exactly that and returns products that match, not «similar» ones. This sounds easy but it's much harder to get right than the market promises: most chatbots fall short the moment the customer combines multiple specific filters (size, color, stock, price range). We explain it in depth in why your AI chatbot can't find products in your catalog.
• No per-message markup on top of Meta's cost. Whatever Meta charges is what you pay; we don't add our own margin on outbound messages.
• Multilingual out of the box. If someone writes to you in German or French, the bot detects the language and replies in it. More in our multilingual chatbot guide.
• Free 7-day PRO trial including WhatsApp. No credit card at signup. If you don't love it, cancel in one click.

In summary

In 2026, building a WhatsApp chatbot is no longer an experiment. Meta's official API is accessible to any business, customer-initiated conversations are free and unlimited since July 2025, and there are platforms with real conversational AI that work well without spending a fortune. The question is no longer «is it free?» but «what do I need this bot to do, and how much does what I need cost?».

If you only want to serve inbound (FAQs, catalogue search, contact capture), a platform with an official Meta connection like ours gets you live in minutes for $23-59/month. If you're also going to push outbound marketing with templates, you're looking at $50-150/month and a platform with a sending module. If your volume is enterprise, the floor is around $500/month with custom seats.

The market traps are the usual ones: vendors who call something an «autonomous AI agent» when it's actually button flows with an AI-generated response inside, platforms that charge a 25% markup on every Meta template without telling you, or «free plans» that don't actually cover WhatsApp. The way to avoid them is the same as always: ask for a demo with your case, read the small print on per-message pricing, and test before you pay.

When you're ready to start, you can try the Bravos AI WhatsApp chatbot for 7 days on the PRO plan, free, no credit card at signup, no commitment.

Frequently asked questions

What is a WhatsApp chatbot?

It's a program that automatically replies to the messages sent to your WhatsApp Business number, with no one typing on the other side. If it has AI behind it, it understands what the customer asks in plain language and replies with your business information (website, catalogue, PDFs, hours). It's different from the fixed greeting and away messages of the WhatsApp Business app.

What's the difference between WhatsApp Business and WhatsApp Business API?

WhatsApp Business is the free mobile app a business downloads to have a number separate from the personal one, with a basic catalogue and fixed messages. WhatsApp Business API is the technical channel Meta offers so external software (like an AI chatbot) can send and receive messages on your number. The API has no app and is operated from the dashboard of whatever platform you use.

What's the best free WhatsApp chatbot?

A conversational chatbot with real AI that's free forever doesn't exist. What's 100% free is the fixed greeting and away message of the official WhatsApp Business app, no AI. If you want a chatbot that understands the customer and replies with real business data, you're looking at $23-200/month depending on the platform. Many, ours included, offer a free 7-14 day trial so you can see if it fits before paying.

How do I create a WhatsApp chatbot?

In short: decide what tasks you want it to do, gather your business information (website, PDFs, catalogue, hours), sign up to a platform, upload that info to train it, connect your WhatsApp number through Meta Business signup, and test before going live. For a chatbot that only replies to inbound, you don't need to pass Meta business verification. The full guide is in how to build a WhatsApp chatbot step by step.

How much does an AI WhatsApp chatbot cost?

It depends on the case. For inbound only (real AI, multichannel web + WhatsApp, contact capture), from $23/month (Bravos AI Starter and PRO at $59). If you also want to send outbound marketing templates, $50-150/month (Tidio with Lyro, ManyChat, Wati, Landbot). For mid-sized companies with many messaging apps connected at once, $100-300/month (Respond.io, Interakt). Enterprise is custom-priced from around $500/month (Bravos AI Enterprise, Intercom Fin, HubSpot). On top of that is Meta's cost per outbound marketing message (~$0.025 in the US); replies to customers who message first are free since July 2025.

Can ChatGPT be integrated with WhatsApp?

Yes, but not «ChatGPT» the app, rather the underlying OpenAI model (GPT-4, GPT-4.1, etc.) via its API. Most modern platforms, Bravos AI included, use that model (or equivalents like Claude or Gemini) under the hood. What the customer sees is a chatbot that understands well, not «ChatGPT on WhatsApp». If you want to build it yourself with OpenAI's API plus a tool like n8n, it's possible but it'll take 13-26 hours the first time. We compared it here.

What is WhatsApp Business API?

It's the set of features Meta offers businesses to connect their WhatsApp number with external software: chatbots, customer service dashboards, CRMs. It has no mobile app of its own; it lives inside whichever platform decides to use it. Its two key features are message templates (approved text for the business to start a conversation with a customer) and the 24-hour window (during which your chatbot replies freely with no cost after a customer message). For a chatbot that only serves inbound, you don't need to pass Meta business verification.

Sources

• Meta — WhatsApp Business Platform
• Meta — WhatsApp Cloud API overview
• Meta — WhatsApp Business pricing (July 2025 update)
• Meta — WhatsApp Commerce Policy
• Meta — WhatsApp Cloud API launch (May 2022)