A chatbot is a program that talks with people through text, usually in a chat window on a website, in a messaging app or inside a product. Some follow a fixed script of buttons and canned answers. Newer ones use an AI language model to understand free-form questions and write their own replies.
Scripted bots versus AI bots
Chatbots are older than most people think: ELIZA, built at MIT in the mid-1960s, mimicked a therapist by matching patterns in what users typed. Rule-based bots still work the same basic way. Someone designs a decision tree ("Track an order" leads to "Enter your order number") and the bot can only handle paths that were planned.
AI chatbots send the conversation to an LLM, which writes a reply. On their own these models know nothing about your business, so good business bots add your content using RAG, and some connect to systems like a booking calendar so they can act, which moves them toward being an AI agent. On a website the bot usually appears as a small widget in the corner, added with a snippet of script.
A practical example: a dental clinic
A clinic gets the same questions every day: opening hours, whether it accepts a certain insurer, how to reschedule. A chatbot trained on the clinic's own FAQ answers those at night and on weekends, offers a link to the booking page, and hands anything medical or unclear to staff by collecting a name and phone number. The front desk spends less time on the phone, and patients get answers outside office hours.
Should your site have one?
A chatbot earns its place when you receive many repetitive questions or want to capture leads after hours. If you get a few enquiries a week, a clear FAQ page and a simple contact form usually serve visitors better and cost nothing to maintain. A bot that cannot answer is worse than no bot.
Chatbot mistakes to avoid
- No route to a human. Always offer email, phone or a handoff when the bot is stuck.
- Letting it invent policies. Restrict an AI bot to your documents and tell it to admit when it does not know.
- Popping open instantly. An unrequested chat window covering content annoys visitors, especially on phones.
- Never reading the transcripts. Conversation logs show what customers want and where the bot fails.
- Ignoring privacy rules. If the bot collects personal details, say so in your privacy policy.
Related terms
- LLM: the model behind AI chatbots.
- RAG: how a bot answers from your own content.
- AI agent: a bot that can also take actions.
- Lead generation: a common job for website bots.