A WhatsApp chatbot is a system that automatically talks to customers through a company’s number, answering questions and guiding them through service and sales steps. For professional use, the safest path is an AI chatbot connected to the WhatsApp Business API, with a clear handoff to a human agent.
This guide explains the types of bots, the official connection, coexistence, what is free, costs from October 2026, how to build one step by step, and which metrics to track.
Key Takeaways
- Menu chatbots handle simple scripts, AI chatbots understand free-form language, and AI agents also query systems and take actions.
- The official API with coexistence keeps the number and the app in use, with lower risk than unofficial QR code connections.
- From October 1, 2026, Meta charges for API service messages above 1,000 free per phone number per month, while messages sent from the app under coexistence stay free.
What Is a WhatsApp Chatbot?
The word chatbot covers very different tools. Understanding the three most common generations helps a business compare proposals by what each one actually delivers, not by the product name on a vendor’s website.
Definition and Where the Chatbot Works
A WhatsApp chatbot is software connected to the business number that receives customer messages, interprets the request, and replies without an agent typing. It answers questions, sends documents, records orders, and routes the conversation to the team when needed.
In practice, it works as the first layer of service: it handles the repetitive work and prepares the ground for a human in cases that require negotiation, empathy, or a decision. The step-by-step process for building an AI agent, from knowledge base to testing, is in the guide on how to create AI agents for WhatsApp.
Flow Chatbot, AI Chatbot, and AI Agent
The three categories differ in what the system understands and what it can do once it understands:
- Flow chatbot (menu): follows a tree of fixed options, such as “press 1 for support.” It is predictable and easy to maintain, but it stalls when the customer writes something off script.
- AI chatbot: uses language models to understand freely written questions, typos included, and replies based on content defined by the business.
- AI agent: besides chatting, takes actions in other systems, such as checking inventory, creating a lead in the CRM, or booking a slot on the calendar.
The practical difference shows up at the end of the conversation. A menu bot delivers information, while an AI agent delivers a recorded result, such as an order created or an appointment confirmed. This model is covered in the article on AI agent CRM integration.
When Each Model Makes Sense
Menu flows still work well for short, stable scripts, such as sending an invoice copy or choosing a department. When questions vary a lot, an AI chatbot reduces drop-off, because the customer does not have to guess which option matches their problem.
An AI agent makes more sense when the conversation ends in a transaction: booking, selling, collecting a payment, or updating a record. Many companies combine all three, with buttons for clear-cut choices and AI for free text, as the comparison of AI models for chatbots shows.
A business can also connect a general-purpose language model to its own number, but it needs instructions, limits, and integration with human service. This is different from offering a general-purpose AI assistant to the public: since January 15, 2026, Meta’s terms allow that only where it is legally required to, which today means Brazil (+55), according to Meta’s page on AI providers.
How a Chatbot Works on the Official API
Every professional WhatsApp chatbot depends on a technical connection to the number. The WhatsApp Cloud API, hosted by Meta, is the official entry point and defines what the bot can send, in which format, and at which point in the conversation.
Webhooks, Service Messages, and the 24-Hour Window
On the official API, each incoming message reaches the platform through a webhook, along with delivery statuses for sent messages, according to the platform overview published by Meta. The chatbot reads the event, decides on the reply, and sends it back through the API.
When the customer writes, a 24-hour customer service window opens, and it renews with each new message from them. Inside it, the bot can send any service message. Outside it, only approved templates, under the message sending rules in the official documentation.
Buttons, Lists, and Flows
The API supports much more than text: reply buttons, lists, location requests, documents, audio, images, video, catalog messages, and reactions. These features reduce typing and speed up the conversation for people on their phones.
For multi-step data collection, WhatsApp Flows create forms inside the conversation, useful for booking appointments, generating leads, or gathering feedback. The WhatsApp Flows guide shows how to combine these forms with an AI agent.
Templates to Start Conversations
When the business starts the contact, as with a reminder or an order confirmation, the chatbot must use a template approved by Meta. Each template has a category, such as marketing, utility, or authentication, which defines its usage rules and price.
Proactive messages also require consent. Meta’s policy asks businesses to contact only people who provided their number and gave WhatsApp opt-in to receive messages, and to honor opt-out requests made inside or outside the app.
Unofficial Connection, Official API, and Coexistence
Many low-cost chatbot offers connect the number by QR code, imitating WhatsApp Web. The difference from the official connection is not only technical. It lies in the risk the business takes on for its own number and its customer history.
Why a QR Code Connection Puts the Number at Risk
The WhatsApp Business Terms of Service prohibit developing or using applications that interact with the services without prior written consent. In case of violation, the account can be limited, suspended, or terminated, and the business can be barred from creating another one without authorization.
Unofficial WhatsApp API solutions depend on this type of connection and can stop working after an app update. No path removes all risk, but the official API greatly reduces exposure. Anyone who has already lost access can find guidance in the article on a banned WhatsApp number.
Coexistence: App and API on the Same Number
WhatsApp coexistence connects the WhatsApp Business app account to the Cloud API through a Meta partner. The number stays the same, the team keeps replying from the phone, and the chatbot answers through the API, with conversations mirrored on both sides.
According to the coexistence documentation, activation imports the last 180 days of history, and groups, calls, and the catalog keep working in the app. Messages sent from the app are not charged, and the primary phone must be used regularly so the connection does not drop.
AI Chatbots for WhatsApp in Service and Sales
AI chatbots perform best when they are designed for a clear goal. The uses below account for most AI WhatsApp chatbot projects and can be switched on gradually, starting with the team’s most repetitive task.
Customer Service Chatbot on WhatsApp
In support, the chatbot answers frequent questions and shares hours, order status, and return policies at any time. With a well-written knowledge base, the AI handles variations of the same question without making the customer navigate menus.
The critical point is the handoff to a human. The WhatsApp Business Messaging Policy allows automation, as long as there is a prompt, clear, and direct path to talk to a person. The hybrid model is covered in human and AI customer service.
Sales Chatbot and Lead Qualification
In sales, the chatbot asks the questions a salesperson would ask on first contact: interest, timeline, budget, and location. Based on the answers, it scores the lead and passes the ready-to-talk contacts to the sales team, with a summary already recorded.
This lead qualification process is especially useful for businesses that receive contacts from click-to-WhatsApp ads, because an instant reply keeps prospects from losing interest while they wait for a salesperson.
Scheduling, After-Sales, and Collections
An AI agent can check the calendar, offer open slots, and confirm the booking without human intervention. In after-sales, it sends usage instructions, tracking codes, and review requests.
In collections, utility templates remind customers of due dates and send the payment link, while the chatbot negotiates dates within defined rules. WhatsApp sales follow-up follows the same logic of timely, permission-based contact. WhatsApp’s policy prohibits asking for full card numbers in the conversation.
Which Businesses Benefit Most From a Chatbot
A chatbot pays off when WhatsApp already concentrates a meaningful volume of repetitive conversations, or when slow replies cost sales. Company size changes the setup and the budget, but not the logic of starting with the most common questions.
Small Businesses and Local Shops
For small businesses, the biggest gain is usually replying after hours and freeing the owner from repeated questions about prices, address, and availability. A lean setup focused on these few topics is often enough to start.
In these cases, coexistence carries special weight: the owner keeps using the familiar phone and adds automation without changing numbers or losing the history with existing customers. A small team gains time without changing its routine all at once.
Clinics, E-Commerce, Real Estate, and Food Delivery
Industries with recurring appointments or orders benefit from agents that take actions. Clinics confirm and reschedule appointments, e-commerce stores handle delivery status and returns, and real estate agencies qualify prospects before booking a viewing.
In food delivery on WhatsApp, the chatbot presents the menu, builds the order, and shares the estimated delivery time. In every industry, the design starts from the real questions the team answers today, not from a generic script copied from another business.
Free WhatsApp Chatbot: What You Can Get at No Cost
Searches for a free WhatsApp chatbot are common, and the honest answer has two parts. The app offers simple automations at no cost, and the API has a monthly allowance of free service messages. A complete AI chatbot, however, always involves some cost.
Free Features in the WhatsApp Business App
The WhatsApp Business app is free and designed for small businesses, with one number per account, according to the official comparison between WhatsApp and WhatsApp Business. The tools listed on the app features page include:
- Greeting message: an automatic welcome when a new customer starts a conversation.
- Away message: an automatic reply to messages received outside business hours.
- Quick replies: shortcuts for frequent messages, sent by the agent.
- Labels: organize and filter conversations that need attention.
- Catalog: a showcase of products and services inside the app.
These features cover the basics, but they do not interpret questions or take actions. Quick replies, for example, depend on someone choosing and sending each message, and the greeting sends the same text to every customer.
Free Service Message Allowance on the API
On the official API, Meta charges per delivered message, based on the category and the recipient’s country, as shown on the platform’s pricing page. From October 1, 2026, Meta also charges for service messages at the market’s utility and authentication rate. In Brazil, for example, that is R$ 0.035 per delivered message, compared with R$ 0.3217 for marketing.
According to the page on non-template messages, each phone number gets 1,000 free service messages delivered per month. For low-volume businesses, this can cover many of the bot’s replies, not counting platform and AI costs.
The Limits of Free and When to Pay
Free tools usually limit contacts, agents, or integrations, and some use unofficial connections. The app was also not built for large teams: the official comparison points businesses that need multiple agents and bots integrated with other systems to the platform.
Paying makes sense when conversation volume exceeds what one person can answer well, when slow replies lose sales, or when the bot needs to check the calendar, inventory, and CRM. At that point, the comparison is no longer free versus paid. It becomes automation versus lost opportunities.
How Much Does a WhatsApp Chatbot Cost?
The total cost of a WhatsApp chatbot adds up three parts: Meta’s fees, the platform subscription, and AI usage. Only the first is publicly priced, and from October 2026 it includes service messages for businesses that reply through the API.
Meta’s Fees From October 2026
Marketing, utility outside the window, and authentication are charged on delivery. From October 1, 2026, Meta also charges for service messages above the 1,000 free per phone number each month and for utility templates sent inside the window. The change is summarized in the WhatsApp API pricing guide.
There are two important exceptions. When the business replies within 24 hours to a contact that came from a click-to-WhatsApp ad, a free 72-hour window opens. Under coexistence, messages sent from the app remain free.
For businesses messaging Brazilian numbers, the category values can be simulated in the cost calculator (Brazil, in Portuguese), which helps estimate the monthly spend from the expected volume of conversations and campaigns.
Platform, AI, and Implementation Costs
The platform charges for the service tool, the agents, and the integrations, and each vendor uses its own model: per user, per conversation, or per package. For AI, usage varies with the length of the replies and of the content consulted.
Team time also counts: writing the knowledge base, reviewing replies, and monitoring the first days. To compare proposals, a business should estimate the cost per resolved conversation, including Meta’s fees, not just the monthly subscription.
How to Build an AI WhatsApp Chatbot for Your Business
Building a WhatsApp Business chatbot on the API does not require coding when the platform offers an agent builder. What defines the result is the preparation: goal, content, integrations, and rules for handing off to an agent.
Preparation: Goal, Content, and Number
Before configuring any tool, it pays to read recent conversations and list the most repeated questions, the tasks that take the most team time, and the moments when customers give up. That list becomes the bot’s initial scope.
On the technical side, the business needs a Meta Business Portfolio, a number that can receive an SMS or call, and an approved display name. Businesses already using the app can go through coexistence. Technical terms are explained in the WhatsApp API glossary.
Step-by-Step Implementation
Once the preparation is done, implementing a WhatsApp chatbot usually follows this order:
- Connect the number: through Meta’s embedded signup on a partner platform, with a new number or through coexistence.
- Create the agent: define the name, tone of voice, limits on what it can answer, and behavior instructions.
- Load the knowledge base: policies, prices, hours, and answers approved by the team.
- Connect the actions: integrate the calendar, CRM, or catalog so the agent can look up and record data.
- Define the human handoff: triggers such as an explicit request, a complaint, or a question with no answer.
- Test and launch: simulate difficult conversations, fix the gaps, and release the bot gradually.
The help center shows how to create an AI agent, and the article on AI agents for WhatsApp goes deeper into writing instructions and configuring tools.
Checklist for Choosing a Platform
When comparing vendors, a few criteria separate a professional solution from an improvised one:
- Official connection: signup through Meta’s screen, with no WhatsApp Web QR code.
- Coexistence: the option to keep the number and the app in use.
- Agent with actions: integration with the CRM, calendar, and in-house systems.
- Shared inbox: multiple agents on the same number, with handoffs between bot and human.
- Templates and campaigns: sending with opt-in and opt-out control.
- Cost transparency: service reports and visibility into Meta’s fees.
- Portability: the ability to switch partners without losing the number.
A fuller analysis of these criteria is in the guide to WhatsApp AI tools, which also compares pricing models and support levels.
Common Mistakes and Metrics to Track
Most chatbot problems do not come from the technology. They come from the conversation design and from a lack of monitoring after launch. Fixing these points early keeps customers from associating the bot with a bad experience.
Mistakes That Hurt the Experience
The most frequent slips show up in the first weeks of operation:
- Hiding the human agent: it irritates the customer and goes against the requirement for a clear path to a person.
- Outdated knowledge base: wrong prices and dates produce confident, incorrect answers.
- Messages without opt-in: blocks and reports damage the number’s reputation.
- Too much AI autonomy: the agent should take only the planned actions, with permissions validated in the system.
Blocks and reports lower the number’s WhatsApp quality rating and can limit sending. For AI, watch out for prompt injection, when someone tries to manipulate the agent with hidden instructions. OWASP recommends least privilege and human approval for risky actions.
Metrics to Track
A simple dashboard, reviewed every week, shows whether the bot is resolving issues or just delaying service:
- Resolution rate without a human: conversations closed by the bot with the problem solved.
- Handoff rate: how many conversations go to the team, and why.
- Time to first response: it should be instant, at any time of day.
- Conversion: qualified leads, bookings, or orders generated by the conversation.
- Satisfaction (CSAT): the score the customer gives at the end of the interaction.
Reviewing handed-off conversations shows where the knowledge base falls short. Under Brazil’s data protection law (LGPD, in Portuguese), Article 20 lets individuals request a review of decisions made solely through automated processing, which reinforces the value of logs and human oversight. Similar rules apply under the GDPR in the EU.
WhatsApp Chatbot FAQ
No, as long as the platform offers an agent builder. Setup involves text, rules, and ready-made integrations, and coding only comes in for custom integrations with the company’s own systems.
Yes, as long as it uses templates approved by Meta and sends only to contacts who opted in. Each delivered template is charged by category and recipient country, and opt-out requests must be honored.
It can, if it lacks clear limits. The risk drops when the agent answers only from content approved by the business, admits when it does not know the answer, and hands the conversation to a human agent in those cases.
The official API receives audio, images, video, and documents. Whether voice notes are transcribed or images interpreted depends on the AI features of the chosen platform, so it is worth confirming this before signing up.
On the official API, yes. Meta allows a number to be migrated between partners while keeping the display name, quality rating, and messaging limits, and templates are copied over. This keeps the business from being locked into a single vendor.
No. It takes over repetitive questions and standardized tasks, while the team handles negotiations, complaints, and sensitive cases. WhatsApp’s policy even requires a clear path for the customer to talk to a person.
Conclusion
A WhatsApp chatbot delivers results when it combines an official connection, AI fed with the company’s content, and a simple handoff to a person. Free features help at the start, but service at scale, sales, and scheduling call for an integrated agent and close cost tracking.
ConverZap is a Brazilian platform for AI agents and WhatsApp automation on Meta’s official API, with coexistence with the app. To design the right chatbot for your operation, talk to the ConverZap team.
