⚡ A chatbot is software that simulates conversation with humans through text or voice. Early chatbots followed rigid scripts and broke the moment you went off-script. Modern AI chatbots — powered by large language models — understand natural language, handle questions they have never seen before, and maintain context across a conversation the way a person would.

Category: Generative AI · Difficulty: Beginner · Last updated: 15 May 2026 · 5 min read

What is Chatbot?

The first chatbot, ELIZA, was built at MIT in 1966. It could hold a surprisingly convincing conversation — but only by pattern-matching your words and reflecting them back with pre-written responses. Ask it something outside its script and it fell apart immediately. For fifty years, most chatbots stayed close to this model: decision trees, keyword matching, scripted flows.

Then large language models arrived and everything changed. ChatGPT is a chatbot — but it is a chatbot built on a model trained on trillions of words of human text, capable of understanding the intent behind any question, maintaining context across a long conversation, and generating responses it has never produced before. The gap between ELIZA and ChatGPT is the gap between a phone book and a librarian.

How Chatbot works?

Rule-based chatbot:

  1. User sends a message.
  2. The bot matches keywords or patterns against a fixed decision tree.
  3. It returns the pre-written response mapped to that pattern.
  4. If no pattern matches, it falls back to a default (“I didn’t understand that, please try again”).

AI chatbot (LLM-powered):

  1. User sends a message — the full conversation history is included as context.
  2. The LLM processes the entire context and generates a response token by token.
  3. The response is shaped by the system prompt, conversation history, and any tools the bot has access to (search, databases, APIs).
  4. The conversation continues, with history growing until the context window limit is reached.

Real-world examples

Not theory — what real teams actually shipped using this technique.

  • Klarna’s AI chatbot handles 2.3 million customer service conversations per month autonomously — resolving disputes, processing returns, and answering billing questions without human agents.
  • Woebot is a mental health chatbot that uses CBT (Cognitive Behavioural Therapy) techniques to help users manage anxiety and depression — available 24/7 at a fraction of the cost of human therapy.
  • Duolingo’s AI chatbot lets language learners practice real conversation in a second language — responding naturally to any sentence the learner types and correcting grammar in context.

Common pitfalls

  • Hallucination — AI chatbots can confidently state incorrect information. For customer support, always ground the bot in verified knowledge bases and add fact-checking layers.
  • Context window exhaustion — very long conversations degrade as early context falls out of the model’s window. Summarise conversation history periodically in long-running sessions.
  • Escalation failures — bots that cannot recognise when a user is frustrated, confused, or needs a human create terrible experiences. Always design a clear escalation path to a human agent.
  • Over-automation — not every customer interaction should be handled by a bot. High-emotion situations (complaints, cancellations, grief) typically need human empathy that no current chatbot replicates reliably.

Frequently asked questions

QUESTION 1 What is a chatbot in simple terms?

ANSWER 1 Software that has a conversation with you — from simple FAQ bots that match questions to fixed answers, to advanced systems like ChatGPT that understand open-ended questions and respond as flexibly as a human.

QUESTION 2 What is the difference between a rule-based and AI chatbot?

ANSWER 2 Rule-based follows a script and breaks outside it. AI chatbots use language models to understand any message and respond flexibly, maintaining context across the conversation.

QUESTION 3 What are chatbots used for?

ANSWER 3 Customer support, e-commerce, healthcare symptom checking, HR policy questions, and education tutoring.

QUESTION 4 What are the limitations of chatbots?

ANSWER 4 Rule-based bots fail outside scripts. AI chatbots can hallucinate. Neither handles genuine emotional complexity well. Long conversations degrade as context falls out of the window.

📬 Get one concept + one use case every Tuesday. Join the newsletter →