A chatbot is a software program that holds a conversation with people in text or voice, on a website, in a messaging app or on the phone. It reads what someone types or says, works out what they want, and replies from a fixed script (a rule-based chatbot) or with an AI model that understands language (an AI chatbot).
That is the whole idea. The rest of this guide explains the four types of chatbots you will meet, how a chatbot works step by step, what real chatbots look like, how a chatbot differs from conversational AI, and what a chatbot can do for a business. If you are choosing one, there is a short quiz near the end that points you to the type that fits.
What is a chatbot?
A chatbot is a computer program you talk to. The word joins "chat" and "robot", and the job is the same whatever is under the hood: take a message from a person, decide what it means, and send back a useful reply without a human typing it.
What changes from one chatbot to another is how it decides what to say. Older chatbots follow rules someone wrote by hand: if the message contains "opening hours", show the opening hours. Modern AI chatbots use natural language processing, the branch of AI that lets software work out the structure and meaning of text (Google Cloud), and large language models that write a fresh answer every time. If you want the AI side in depth, read what is an AI chatbot.
In everyday use, "chatbot" covers all of these: the menu of buttons on a delivery site, the assistant on a bank app, a support agent on a shop website, and general-purpose AI assistants such as ChatGPT, Claude and Gemini.
What are the four types of chatbots?
There is no single official list, but most guides group chatbots into four types by how they choose a reply:
- Menu or button-based chatbots. The visitor clicks options instead of typing. Each click opens the next branch of a decision tree.
- Rule-based (keyword) chatbots. The visitor types, and the bot looks for keywords or patterns it was given, then returns the answer tied to that rule. Anything outside the rules gets a fallback such as "Sorry, I did not understand".
- AI chatbots. The bot uses a language model to understand intent, however the question is worded, and writes an answer. Business AI chatbots are usually trained on the company's own website, documents and FAQs, so they answer from that content.
- Hybrid chatbots. Fixed flows for the few tasks that must follow exact steps, such as a returns form, with an AI chatbot answering everything else, and a route to a person when needed.
| Type | How it picks a reply | Handles new wording? | Upkeep | Best for |
|---|---|---|---|---|
| Menu or button-based | The visitor clicks through a decision tree | No typing, so no | Edit the tree by hand | A handful of fixed choices, such as tracking an order |
| Rule-based (keyword) | Matches keywords or patterns to scripted answers | Only if the keywords appear | Add rules for every new question | Small, predictable sets of questions |
| AI chatbot | A language model reads intent and writes an answer from its sources | Yes | Keep the source content current | Varied questions that change over time |
| Hybrid | Fixed flows for set tasks, AI for open questions, people for the rest | Yes, outside the fixed flows | Both, but each part stays small | Businesses with a few strict processes and many open questions |
For a deeper side-by-side on cost and features, see our rule-based vs AI chatbot buyer's guide.
How does a chatbot work?
Every chatbot runs the same loop. What differs is how clever each step is.
- Receive the message. The visitor types, taps a button or speaks. A voice chatbot first turns speech into text.
- Work out what the person wants. A rule-based bot checks its keyword list. An AI chatbot uses a language model to read the intent and the context of the whole conversation, so "can I come Saturday instead?" makes sense after an earlier message about a booking.
- Find the answer. A scripted bot pulls the reply tied to the matched rule. A business AI chatbot usually searches its knowledge (your web pages, files and FAQs) for the passages that answer the question, then asks the model to answer from them. This pattern is called retrieval-augmented generation, or RAG, and it grounds the reply in your own information instead of the model's general memory (Google Cloud).
- Reply. The bot sends the answer in the same channel, in text or voice.
- Hand over when it should. A good chatbot knows its limits: when the content does not cover a question, or the visitor asks for a person, it passes the conversation to a human or takes contact details.
For the full walk-through with a worked conversation, read how AI chatbots work, step by step.
What is an example of a chatbot?
- A website support chatbot. The chat bubble in the corner of a company site that answers questions about products, delivery, returns or opening hours.
- A booking chatbot. A clinic, salon or restaurant bot that answers questions and helps a visitor find or request a time.
- A messaging app chatbot. A business account on WhatsApp or another messaging app that replies to customers inside the app they already use.
- A voice assistant. A phone line or smart speaker that understands spoken questions and answers out loud.
- A general-purpose AI assistant. ChatGPT, Claude and Gemini are chatbots too. They answer from broad general knowledge, while a business chatbot answers from one company's own content.
One question, three chatbots: a worked example
Picture a dental clinic. A patient opens the chat at 9pm and types: "I have a cleaning on Thursday but I need to move it to the weekend, are you open Saturdays?"
- Menu bot: it cannot read the message. It shows four buttons: Book, Opening hours, Prices, Contact. The patient clicks Opening hours, sees the times, then has to start again to change the booking.
- Rule-based bot: it spots the keyword "open" and replies with the full opening hours. It misses the real request, moving an existing appointment.
- AI chatbot trained on the clinic's site: it reads both parts of the message. It answers from the clinic's hours page that the clinic is open on Saturday mornings, explains the clinic's own steps for moving an appointment, and offers to pass the request to the front desk with the patient's name and email so someone confirms the new time.
The AI chatbot does not invent a Saturday slot or a policy. It answers from what the clinic published and hands the booking change to a person. That combination, accurate answers from your own content plus a clean handover, is what most businesses want from a chatbot today. Try the quiz below to see which type fits your own questions.
Chatbot vs conversational AI: what is the difference?
Conversational AI is the technology. Google Cloud describes it as a type of artificial intelligence that can simulate human conversation, made possible by natural language processing and foundation models (Google Cloud). A chatbot is a product: the thing on the screen that people talk to.
So an AI chatbot is built with conversational AI, but not every chatbot is. A menu or keyword bot is a chatbot with no conversational AI inside. And conversational AI also powers things that are not chat windows, such as voice agents on a phone line. Our guide to chatbot vs conversational AI goes further.
You will also hear "AI agent". Google Cloud places simple bots at the least autonomous end, typically following pre-programmed rules, and AI agents at the other end, pursuing goals and completing tasks for users (Google Cloud). Many modern business chatbots sit in between: they answer questions and also take actions such as capturing a lead or handing a chat to a person.
What can a chatbot do for a business?
- Answer repeat questions at any hour. Prices, delivery, opening hours, how-to questions and policies, answered in seconds, including at night and at weekends.
- Capture leads. Ask for a name and email when a visitor shows buying interest, so sales can follow up.
- Route the rest to people. Pass billing problems, complaints or anything the content does not cover to your team, with the conversation attached.
- Show you what customers ask. Transcripts reveal the questions your website does not answer yet.
The purpose behind all of these is the same: answer customers faster without growing the team at the same pace. Our guide to the purpose of a chatbot covers each use with examples.
With Dante AI you build an AI chatbot from your own content: website pages it crawls, file uploads such as PDF and Word, pasted text, and question and answer pairs. Dante AI runs frontier models from OpenAI, Anthropic and Google on paid plans, and you choose the model for each agent. When your content does not cover a question, the agent says it does not have that information rather than guessing. The Free plan costs $0 and includes 100 credits a month, enough to try it on your own site.
Want to see a chatbot answer from your own website? Get started free at dante-ai.com.
Sources
- Google Cloud, Conversational AI (definition of conversational AI), read 27 October 2026.
- Google Cloud, What are AI agents? (agents, assistants and bots compared by autonomy), read 27 October 2026.
- Google Cloud, What is Retrieval-Augmented Generation (RAG)? (grounding answers in retrieved data), read 27 October 2026.
- Google Cloud, What is Natural Language Processing?, read 27 October 2026.
- Dante AI Help Center, Knowledge sources, AI models and Plans and pricing, read 27 October 2026.