What is a chatbot?

A chatbot is a software application that simulates conversation with people through text or voice, answering questions and handling simple tasks without a human agent. Chatbots range from rule-based systems that follow scripted decision trees to AI-powered bots that interpret natural language and generate their own responses.

The word covers a wide spectrum. A menu of tap-to-choose buttons on a retailer's help page is a chatbot. So is a system that understands a typed sentence, looks up an order, and answers in fluent language. What they share is the basic premise: software on one side of the conversation, a person on the other.

This page covers what a chatbot is, the main types, how chatbots are used in customer service, how they differ from conversational AI and AI agents, and where they fall short.

A chatbot in one sentence

A chatbot is software that holds a conversation with a customer — anywhere from a scripted menu to AI that understands what you actually mean.

How a chatbot works

Chatbots fall into two broad families, and the difference determines almost everything about how they behave.

Rule-based chatbots follow a predefined script. They match keywords or present buttons, then route the customer down a branch of a decision tree the designers built in advance. Ask something the tree anticipated and the bot answers cleanly. Ask something it didn't and the bot loops, apologizes, or dead-ends. These systems are predictable and inexpensive, and they work well for narrow, high-frequency questions.

AI chatbots interpret meaning instead of matching keywords. They use natural language processing (NLP) to parse intent, and the current generation is powered by large language models that can handle slang, typos, incomplete sentences, and topic shifts. Rather than retrieving a pre-written string, an AI chatbot generates a response calibrated to what was actually asked.

The practical line between them: a rule-based chatbot can only handle the paths its designers mapped; an AI chatbot can handle paths they never anticipated.

Types of chatbots

Most chatbots in use today fall into one of these categories:

In practice, many production chatbots blend types — a rule-based skeleton for sensitive workflows with an AI layer for everything else.

Chatbot vs. conversational AI vs. AI agent

These three terms get used interchangeably, but they describe different things, and the distinction matters when you're evaluating tools.

A chatbot is the surface a customer touches. Conversational AI is the technology that lets a chatbot understand natural language. An AI agent goes further — it doesn't just answer, it acts. A chatbot can tell a customer how to start a return; an AI agent can generate the label, update the order, and issue the credit.

Chatbots in customer service

Chatbots earn their keep on a few well-defined jobs:

Answering routine questions. Order status, return windows, store hours, password resets — high-volume, low-complexity questions a chatbot can resolve instantly, at any hour, without queue time.

Collecting context before a handoff. A chatbot can gather a customer's order number, the nature of the issue, and relevant details, then route to the right team with that context attached — so the customer doesn't repeat themselves.

Assisting human agents. Behind the scenes, a chatbot can surface knowledge base articles, suggest responses, and summarize a customer's history while a human stays in control of the conversation.

Extending coverage. Chatbots operate outside business hours and absorb volume spikes that would otherwise mean long waits.

Handled well, a chatbot resolves the predictable questions so people can focus on the conversations that genuinely need them. That's real value — and it's the floor, not the ceiling. A chatbot focused only on resolving individual requests answers the question in front of it. Whether it can recognize a first-time shopper versus a decade-long customer depends on the customer context available to it. The more useful bar is whether the experience carries that context across channels and can actually resolve the issue rather than simply route it.

Limitations of chatbots

Rule-based bots break on anything unexpected. Off-script input produces loops, dead-ends, or the dreaded "I didn't understand that." The narrower the script, the more often customers hit the wall.

Without customer context, every chat starts from zero. A chatbot reading only the current message can't account for prior contacts, loyalty, or stated preferences. Connected to a unified customer record, it can; on its own, it treats everyone like a stranger.

AI chatbots can hallucinate. LLM-powered bots can state a policy, price, or process confidently and incorrectly. Grounding responses against a curated knowledge base and keeping humans in the loop on high-stakes interactions is the practical mitigation.

Bad handoffs undo the benefit. When a chatbot passes a customer to a human without the conversation history, the customer starts over — and the chatbot becomes the thing standing between them and help, rather than a shortcut to it.