chatbot
A chatbot is a program with which you can converse in writing or speech using everyday language, and which answers your questions immediately in a human style.
In plain terms
A chatbot is a conversational program: you type or say what you want, and it responds. You do not need to search through menus or learn commands; it is enough to phrase your request as you would to a friend.
It is like a customer service desk where someone sits around the clock. Older chatbots could only read from a pre-written script at that desk: if your question was not in it, they got stuck. Today's chatbots, built on large language models, phrase answers in their own words, so you can talk to them about almost anything.
It is good for quickly getting an explanation, summary, text draft, translation, or idea, and for handling simple matters at a company without standing in line. It differs from a traditional search engine in that it does not provide a list of results but a ready answer, and you can ask follow-up questions.
How it works
- You enter your question or request. The program receives this along with the previous conversation, allowing it to refer back to what you said earlier.
- A rule-based chatbot searches your text for keywords and returns a pre-written answer for each.
- Behind the modern chatbot stands a large language model. It learned language patterns from vast amounts of text and was then specifically trained to respond as a helpful conversational partner. It assembles the answer in small pieces (word by word, token by token): at each step, it calculates the most probable continuation and selects from these.
- Many chatbots also work from external sources: they search the internet or pull the answer from the company's own documents and databases, so their answers can be more up to date and more accurate.
- The answer appears on the screen, and you can refine it, ask follow-up questions, or steer the conversation in a new direction.
An everyday example
At 10 p.m. it occurs to you that you don't know when your ordered package will arrive. A chat window pops up in the corner of the online store, you enter the order number, and the chatbot tells you within half a minute where the delivery stands. If your question is too complex, it hands you over to a human customer service representative.
Who is affected
A common misconception
Limits and risks
- It can convincingly present fabricated facts, dates, or sources; this is called hallucination.
- What you type goes to the provider's server and, depending on settings, may be used to further develop the model. Therefore, do not carelessly share passwords, health data, or confidential company information with it.
- Its knowledge is valid only up to a specific date, so it knows about recent events only if it uses internet search.
- It is useful for gathering information on medical, legal, and financial matters, but does not replace professional judgment.
Similar, but not the same
- AI-agent – A chatbot answers you, while an agent independently performs the task in multiple steps, such as booking, modifying a file, or running a program.
- large language model – The language model is the engine, while the chatbot is the conversational interface built on top of it that you use.
- voice assistant – Voice assistants traditionally execute short verbal commands, such as playing music or setting a timer, while the main task of a chatbot is conversation itself. The line is increasingly blurring because large language models are also placed behind voice assistants.
Related concepts
artificial intelligence, generative AI, Natural language processing
customer service chatbot, rule-based chatbot, AI partner
large language model, AI-agent, virtual assistant, prompt, hallucination
A brief history
The first known conversational program, ELIZA, was created by Joseph Weizenbaum at MIT and introduced in 1966. Its best-known variant (DOCTOR) mimicked a psychotherapist using simple pattern matching. The turning point came with the launch of ChatGPT in November 2022, which turned chatbots based on large language models into mass-market products.
Where things stand
By the end of 2026, chatbots have become everyday tools: hundreds of millions of people use the most popular services weekly, and they are now standard in corporate customer service. The trend is that chatbots increasingly perform tasks independently, functioning as agents.
In the news
- Year-long trial: Rosie AI chatbot did not reduce anxiety or depression in pregnant women and new mothersNov 6, 2026 · Education & Family
- In a 400-person trial, the Rosie AI chatbot did not reduce anxiety or depression in pregnant women and new mothersNov 6, 2026 · Health
- Child safety lawsuits against Snap, Meta and Google mount; West Virginia also sues SnapNov 5, 2026 · Education & Family
- AWS launches new package to help bring physical AI into manufacturingNov 5, 2026 · Ipar
- Google sued after claims Gemini gave suicide-related content to a teenagerNov 5, 2026 · Health
- Studies: children attached to AI chatbots show markedly worse mental healthNov 4, 2026 · Education & Family
- EU court adviser says Google's AI chatbot may infringe copyright in press contentNov 3, 2026 · Marketing
- Hungarian court makes new EU referral on press publications and LLM chatbotsNov 3, 2026 · Jog
Also known as
chatbot · csetbot · chatbot · chatbot · chatbot · AI-chatbot · MI-chatbot · chatterbot
Sources
This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.