AI glossary Conversational
What is a chatbot?
← All glossary termsA conversational assistant that talks with users in text, driven either by predefined rules or by a large language model.
Explained
A chatbot is software that holds a conversation with users through text, and sometimes voice. Rule-based chatbots recognise an intent from the user's message and reply with predefined answers or flows. LLM-based chatbots use a large language model to understand open-ended questions and generate natural replies, often grounded in company documents through retrieval. They are used in customer service, sales and internal tools, wherever people need answers or simple tasks done without waiting for a person.
Rule-based vs LLM-based
A rule-based chatbot works from a map of intents and answers. It is predictable and easy to audit, but it only handles what was foreseen, and users notice quickly when they step outside the script.
An LLM-based chatbot understands free text and can answer questions nobody wrote a rule for. It needs other controls instead: grounding in approved content to avoid invented answers, guardrails on what it may say, and evaluation of its replies. Many deployments combine both, with fixed flows for sensitive steps and a language model for everything else.
Chatbot, assistant or agent?
A chatbot answers. An assistant answers and helps with tasks inside a given tool, such as drafting a reply or summarising a case. An agent goes further: it plans steps and uses tools or systems to complete a task, such as changing a booking or opening a ticket.
The line between them is blurring as chatbots gain access to tools. What matters for design is what the system is allowed to do, with which permissions, and when a person must step in.
What makes a chatbot work well
- A clear scope: which questions and tasks it covers, and which it hands over.
- Answers grounded in current, approved content, with sources where useful.
- An easy route to a human, with the conversation history passed on.
- Integration with the systems that hold the answer: orders, bookings, accounts.
- Measurement of resolution, not only of conversations started.
Frequently asked questions
What was the first chatbot?
ELIZA, written by Joseph Weizenbaum at MIT and described in 1966 in Communications of the ACM, is usually cited as the first. It matched keywords in the user's sentence and rearranged them into replies, most famously imitating a psychotherapist.
Is ChatGPT a chatbot?
Yes, in the broad sense: it is a conversational interface to a large language model. It goes beyond classic chatbots because it generates open-ended answers rather than choosing from predefined replies, and it can use tools.
What is the difference between a chatbot and an AI agent?
A chatbot's main job is to converse and answer. An AI agent is built to complete tasks: it decides which steps to take and acts on other systems through tools. A chatbot can include agent capabilities, and an agent can have a chat interface.
How do you stop a chatbot from making things up?
Ground it in approved content through retrieval, instruct it to say when it does not know, and limit it to the topics it is meant to cover. Then evaluate its answers regularly with real questions and review the cases it gets wrong.
When should a chatbot hand over to a human?
When the user asks for it, when the request falls outside its scope, when confidence is low, or when the case is sensitive, such as a complaint or a vulnerable customer. The handover should carry the conversation so the user does not have to repeat themselves.
Decision guides that use this term
Where it shows up at Thinkia
Further reading
Related terms
Sources
- Weizenbaum · ELIZA: a computer program for the study of natural language communication between man and machine (Communications of the ACM, 1966)
- Dam et al. · A complete survey on LLM-based AI chatbots (arXiv)
- AWS · What is Amazon Lex V2?
Updated: 2026-10-21