All posts
AI Systems#what is a chatbot#chatbot meaning#what is an AI chatbot

What is a chatbot? The plain answer, the types, and where one stops being enough

What is a chatbot, and what is an AI chatbot? A plain definition, how chatbots work, the types of chatbots from rule based to conversational AI, real chatbot examples, what businesses use them for, their limits, and when a business needs something that can do the work behind the answer.

Md Shohel· October 30, 2026· 13 min read
Listen to this post0:00 / 14:56
AI narration of this post, read in full

A chatbot is software you talk to in writing (or out loud) that answers back. You type a question into a box on a website, in an app or in a messaging tool, and the program replies in plain language. That is the whole definition. Everything else people argue about, rule based or AI, scripted or free, smart or frustrating, is a question about how it produces the reply.

Most people now meet chatbots in two places. One is the little chat bubble in the corner of a company's website. The other is a general assistant like ChatGPT, Claude or Gemini, where the chat is the product. They share a name and very little else, which is why "chatbot" has become a slippery word.

This post covers what a chatbot is, how the different kinds work, real examples, what businesses actually use them for, where they fall short, and the point at which a business needs something that can act on the conversation as well as hold it. We make Sanaf, an AI employee that can sit in a chat on your website and also do the work that comes out of it, so we have a view on that last part. The rest stands on its own.

What is a chatbot?

A chatbot is a computer program that holds a conversation with a person through text or speech. It takes what you say, works out what you mean, and returns an answer. "Chatbot meaning" in one line: a bot you chat with.

The word covers a wide range. At one end sits a menu with buttons ("Track my order", "Talk to someone") wearing a chat window as a costume. At the other end sits a large language model that can write, summarize and reason about almost anything you type. Both are chatbots. They fail in completely different ways, so it helps to know which one you are looking at.

The oldest well known example is ELIZA, written by Joseph Weizenbaum at MIT in the mid 1960s. ELIZA matched keywords in what you typed and reflected them back as questions, in the style of a therapist. It understood nothing. People still confided in it, which tells you something about chatbots that has stayed true for sixty years: a fluent reply feels like understanding, whether or not any is there.

What is an AI chatbot?

An AI chatbot is a chatbot that uses machine learning, and today almost always a large language model, to understand the question and write the answer, instead of following a script somebody drew in advance.

The practical difference shows up the first time you ask something nobody planned for. A scripted bot either matches your words to one of its prepared paths or gives up ("Sorry, I didn't understand that"). An AI chatbot reads the question as written, including typos, follow ups and half sentences, and writes a reply that fits. It can hold the thread of a conversation, so "and what about Saturdays?" makes sense after a question about opening hours.

That flexibility cuts both ways. A scripted bot can only say what it was given. An AI chatbot can say things nobody gave it, including things that are wrong and sound right. The good ones are tied to a business's own material (its help articles, policies, price list) and told to say "I don't know" when the answer isn't there. The approach is usually called retrieval augmented generation, and we wrote a plain English explainer of RAG if you want the mechanics.

How do chatbots work?

Every chatbot runs the same loop, however simple or advanced it is:

  1. Read the message. The person types or speaks, and the bot receives the text (speech gets converted to text first).
  2. Work out the intent. What does this person want? A rule based bot looks for keywords or a button press. An AI chatbot interprets the whole message in context.
  3. Find or produce the answer. A rule based bot picks the prepared reply for that intent. An AI chatbot looks up relevant material, if it is connected to any, and writes a reply.
  4. Reply and wait. The answer goes back, and the loop starts again with the next message.

"What is a chatbot and how does it work" usually comes down to step two and step three. The intent step decides whether the bot understood you. The answer step decides whether what it said was true. When a chatbot frustrates someone, one of those two broke.

There is also an implied step five that most chatbots skip: do something about it. A customer asks to move their appointment. Most chatbots can tell them how to move it, or hand them a link. Very few can move it. Keep that in mind; it comes back later.

Types of chatbots

There are many ways to slice this. For a business deciding what to put in front of customers, four types cover nearly everything on the market.

TypeHow it repliesGood atBreaks when
Rule based chatbotFollows a decision tree of buttons and keywordsNarrow, predictable tasks like order status or opening hoursThe question falls outside the tree
Keyword or intent botMatches the message to a list of trained intentsRouting common questions to set answersPhrasing drifts from the training examples
AI chatbot (LLM based)Writes the reply with a language model, ideally from your own materialOpen questions, follow ups, messy phrasingIt is not grounded and guesses confidently
AI agent or AI employeeUnderstands the request, then acts in your toolsTurning the conversation into finished workIt has no rules for what it may do without asking

Rule based chatbots

A rule based chatbot follows a flowchart. "Are you a new or existing customer?" leads to one branch, and each branch offers the next set of choices. Nothing is generated; every possible reply was written by a person.

Rule based bots are cheap, predictable and easy to check. They never say anything you didn't approve. They are also the reason many people dislike chatbots, because they trap anyone whose question doesn't fit a button. If your customers ask about five things and five things only, a rule based chatbot can still be the right call.

Keyword and intent chatbots

The middle generation tried to understand free text by classifying it into a fixed list of intents: "refund request", "shipping question", "password reset". Each intent had a written answer. They handled more phrasing than a pure decision tree, but anything outside the intent list fell through to a fallback reply. Many website chatbots still run this way under the hood.

AI chatbots and conversational AI

What is conversational AI? It is the umbrella term for software that understands and produces natural language in a back and forth exchange: AI chatbots, voice assistants such as Siri and Alexa, and the language models behind them. An AI chatbot is the text version most people meet. So if you're weighing conversational AI vs chatbot, the honest answer is that one contains the other. All modern AI chatbots are conversational AI; not every chatbot is, since a button tree isn't really conversing.

(People do ask "is Alexa a chatbot?" It is a voice assistant, which is conversational AI. Most people reserve "chatbot" for text, but the line is soft.)

AI agents and AI employees

The fourth type is where the category is heading, and it is a real change in kind. An AI agent understands the request the way an AI chatbot does, and then uses tools to carry it out: it looks up the booking, moves it, and confirms. An AI employee is an agent with a standing job, memory and rules about when to ask first. We go through the full chatbot vs AI agent difference in AI agent vs chatbot, assistant and automation, and the wider family in types of AI agents.

Chatbot examples

Some chatbot examples you have probably used, grouped by type:

  • General AI assistants. ChatGPT, Claude, Gemini and Microsoft Copilot are AI chatbots where the conversation is the product. You ask, they write.
  • Website support bubbles. The chat window on a retailer, airline or software company's site. Older ones are rule based menus; newer ones are AI chatbots trained on the company's help center.
  • Messaging app bots. Bots inside WhatsApp, Facebook Messenger or Slack that answer questions, send updates or run simple workflows.
  • Voice assistants. Siri, Alexa and Google Assistant: conversational AI you talk to rather than type to.
  • Internal help bots. A bot in a company's Slack or Teams that answers "where's the holiday policy?" or "how do I reset my VPN?" from the company's own documents.

AI chatbot examples inside a business tend to look less exciting and more useful. A clinic's site bot that answers "do you take my insurance?" from the clinic's actual list. A software company's bot that pulls the right help article for an error message. A wholesaler's bot that tells a buyer whether an item is in stock. Conversational AI examples like these work because the bot has one narrow body of facts to stand on.

What are chatbots used for?

Most business chatbots do one of a handful of jobs:

  • Answering repeat questions. Hours, prices, policies, delivery times, "do you do X?" The questions a team answers forty times a week.
  • Customer service triage. An AI chatbot for customer service sorts what came in, answers what it can, and passes the rest to a person with the context attached.
  • Capturing leads. A chatbot for a website asks a visitor what they need and collects a name and email, so an interested visitor at 11pm becomes a follow up in the morning.
  • Booking and scheduling. Offering times and taking a booking, when the bot is connected to a calendar.
  • Internal questions. Helping staff find a policy, a process or a document without asking a colleague.

What are the benefits of using AI chatbots?

The benefits of AI chatbots are mostly about time, for both sides of the conversation:

  • Answers outside working hours. A visitor with a question at the weekend gets an answer instead of a contact form and a wait.
  • Instant replies. Nobody waits in a queue for a question that has a two line answer.
  • Fewer repeat questions for the team. The people on the team spend their time on the conversations that need a person.
  • Consistency. A chatbot grounded in your own material gives the same answer on Tuesday as it did on Monday.
  • A record. Every conversation is written down, so you can see what customers actually ask, which is useful even if you never change the bot.

The limits of chatbots

Chatbots have real limits, and it is better to know them before you install one than after a customer finds them for you.

They can be confidently wrong. An AI chatbot that isn't tied to your material will fill gaps with plausible guesses. A refund policy it invented is still a refund policy a customer read on your site. Ground it in your own documents and tell it to say when it doesn't know.

They frustrate people who need a person. Nothing annoys a customer faster than a bot that won't let them reach someone. A good chatbot hands off early, with the conversation attached, so the customer doesn't repeat themselves.

They stop at the answer. This is the big one. Most chatbots can tell a customer what to do and can't do it for them. "To change your booking, log in and go to Account, then Bookings" is an answer. Changing the booking is the work. The gap between the two is where most of the customer's frustration and most of the team's time goes.

What should you do before using an AI chatbot? If you are the one deploying it: decide what it is allowed to answer, give it your real material, test it with the twenty questions customers actually ask, decide how it hands off to a person, and read the first week of conversations. If you are the one typing into one: don't paste in passwords or anything you'd regret being stored, and check anything important it tells you.

Chatbot vs AI agent: when the conversation needs to become work

Look at what customers actually ask for in a website chat. A lot of it isn't a question. It's a request:

  • "Can I move Thursday's appointment to Friday?"
  • "Can you send me a quote for 200 units?"
  • "I paid twice, can you refund one?"
  • "Can someone call me back about the job?"

A chatbot answers all of these with some version of "here's how" or "someone will be in touch". Then a person on the team reads the transcript, opens three apps, and does the thing. The chatbot saved the customer a wait for the first reply, and saved the team nothing on the work.

An AI agent closes that gap. It understands the request and then acts in the tools the business already uses: it checks the calendar and moves the slot, drafts the quote from the price list, finds the duplicate charge. The questions that matter change. With a chatbot you ask "is the answer right?" With an agent you also have to ask "what is it allowed to do, and when does it check with us first?" Sending a quote to the wrong customer is a different kind of mistake from a clumsy sentence.

That is the line between a chatbot and what we'd call an AI customer service agent: one replies, the other resolves, and the second needs rules a chatbot never did.

Where Sanaf fits

Sanaf is an AI employee, hired into your team's Slack or Microsoft Teams. You give it jobs from a roster (inbox, lead follow up, booking, invoicing and so on), and it works inside the apps you already run. On paid plans it can also sit in a chat on your own website, added with one script tag, where it answers visitors from your own material.

What makes it different from a website chatbot is what happens after the visitor's message. When a visitor asks for a booking or a quote, Sanaf can do the work in your connected apps, and it asks your team before anything it cannot undo. The approval arrives as a card in Slack or Teams. When a visitor needs a person, the conversation is handed to your team in Slack with the thread attached, and they reply from the console inbox. Everything it did is written down.

If a chatbot that answers questions is all you need, plenty of good ones exist and you should pick the simplest. If you keep finding that the chat ends and the work starts, that's the gap Sanaf is built for. See what an AI employee is, or check pricing: you start free with $120 of credit and pay for the work, not the seat.

Frequently asked questions

What is a chatbot in simple words?

A chatbot is a program you talk to in a chat window, and it replies in plain language. Some follow a fixed script of buttons; AI chatbots understand free text and write their own replies.

What is the difference between a chatbot and an AI chatbot?

A basic chatbot follows rules somebody wrote in advance and can only give prepared answers. An AI chatbot uses a language model to understand any phrasing and write a fitting reply, which makes it far more flexible and also able to be wrong in ways a scripted bot can't.

What are the main types of chatbots?

Rule based chatbots that follow a decision tree, keyword or intent bots that match messages to set answers, AI chatbots built on language models, and AI agents that act in your tools after they understand the request.

Is ChatGPT a chatbot?

Yes. ChatGPT is an AI chatbot: a general assistant where the conversation is the product. It can also use some tools, which moves parts of it toward agent territory, but for most people it is a chatbot.

What is the difference between conversational AI and a chatbot?

Conversational AI is the broad field of software that understands and produces natural language in a dialogue, including voice assistants. A chatbot is one application of it. Every AI chatbot is conversational AI; a simple button menu chatbot isn't really.

Do I need a chatbot for my website?

If visitors ask the same questions repeatedly, or arrive when nobody is around to reply, a chatbot for your website will save them a wait. If most of what they ask for is a booking, a quote or a change to an order, look for something that can carry the request out, or the chatbot just moves the work to your inbox.

What is the difference between a chatbot and an AI agent?

A chatbot replies. An AI agent replies and then acts: it opens your systems, changes a record, sends the follow up. That makes an agent more useful and also means it needs permissions and approval rules a chatbot never did.

Get the next one by email

We write when we have learned something worth passing on, so nothing arrives unless it is worth your time.

Unsubscribe in one click, any time. Privacy policy