What is an AI chatbot?
An AI chatbot is a program that holds a written conversation in everyday language. It uses a large language model to understand a question and write a reply, often drawing on approved information such as a business's services, hours and policies. Older chatbots followed fixed scripts. AI chatbots handle questions phrased many different ways.
You have used one if you have typed a question into ChatGPT, Gemini or a chat window on a company's website. The difference from older bots is easy to feel. A scripted bot shows buttons and breaks when you type something it did not expect. An AI chatbot reads "my brakes squeal when it rains, is that bad?" and understands it.
How does an AI chatbot work?
An AI chatbot works by reading your message, finding the information it needs, and writing a reply in plain language. A business chatbot adds rules and actions on top. Here is what happens behind one website conversation.
- A visitor types a question, such as "Do you do teeth whitening on Saturdays?"
- The language model works out what the visitor wants: a service and a day.
- The chatbot looks up the business's approved information, often called a knowledge base, for the service list and hours.
- It writes an answer in plain words, based on that information.
- If the visitor wants to go ahead, it takes an action the business allowed: collects a name and number, or offers real openings from the calendar.
- If the question is outside its rules, such as a medical or legal question, it hands off to a person.
Step 3 is what separates a good business chatbot from a risky one. A chatbot that answers only from approved information is less likely to make things up.
How do AI chatbots compare with other ways to chat?
AI chatbots sit between old scripted bots and live chat staff. Here is how the options compare for a local business website.
| Scripted chatbot | AI chatbot | Live chat staff | Contact form | |
|---|---|---|---|---|
| Understands free-typed questions | Rarely | Yes | Yes | Not applicable |
| Available after hours | Yes | Yes | Only if staffed | Yes, but nobody replies |
| Answers right away | Yes, if the question fits a script | Yes | When someone is free | No |
| Can book an appointment | Sometimes, with fixed steps | Yes, when connected to a calendar | Yes | No |
| Risk of a wrong answer | Low, but often unhelpful | Real, if not limited to approved info | Low | None |
Speed matters more than most owners think. Harvard Business Review's 2011 study "The Short Life of Online Sales Leads" found that firms contacting a lead within an hour were nearly 7 times as likely to qualify it. A form that sits in an inbox overnight loses that race.
Where do businesses use AI chatbots?
Businesses use AI chatbots anywhere customers type instead of call. The website chat window is the most familiar. The same technology also answers text messages, Facebook and Instagram messages, and other channels. Larkpin treats website chat and these other channels as separate setups: our AI customer service work covers text and social messages, while a website chatbot lives on your site.
Here are a few examples of what a business chatbot can handle:
- A med spa visitor asks at 11 p.m. which treatments help with sun spots. The chatbot explains the options the spa approved and offers a consultation, without giving medical advice.
- An auto repair shop visitor asks whether the shop works on hybrids and how soon they can bring the car in.
- A law firm visitor describes a car accident. The chatbot collects contact details and basic facts for the firm's intake team and makes clear it cannot give legal advice.
What are the risks of AI chatbots?
The biggest risk of an AI chatbot is a confident wrong answer, and the business is the one held responsible. In Moffatt v. Air Canada (2024), a British Columbia tribunal found the airline liable after its website chatbot gave a customer incorrect information about bereavement fares. The airline argued the chatbot was responsible for its own actions. The tribunal rejected that, finding the chatbot was part of Air Canada's website and the company was responsible for the information on it.
Other risks to plan for:
- Pretending to be human. California's bot disclosure law (SB 1001, in force since July 2019) makes it unlawful to use a bot online with people in California to mislead them about its artificial identity in order to encourage a sale. A clear disclosure that it is a bot avoids that liability.
- Sensitive information. Health and legal details need extra care. For dental, med spa and chiropractic offices, we design setups with HIPAA requirements in mind and keep clinical details out of AI conversations unless your agreements and tools cover them.
- Dead ends. A chatbot that never lets someone reach a person frustrates the people you most want to help.
How can you tell if an AI chatbot is any good?
A good AI chatbot is honest, limited and useful. Before you put one on your site, check these:
- It says it is an AI assistant at the start of the chat.
- It answers only from information you approved, and says "I am not sure" when it should.
- It offers a real person any time someone asks.
- It books only real openings from your live calendar.
- It sends you a summary of every conversation.
- You can read transcripts and change its answers easily.
If a chatbot does more than answer, such as updating records or taking several steps on its own, it starts to become an AI agent. Read what AI agents are for that next step, or what AI automation is for the bigger picture. For phone calls instead of chat, see what a virtual receptionist is.
See what happens when someone messages you after hours
Now you know what an AI chatbot is. The next question is what happens on your own site when a customer reaches out at night. Our Free Missed-Lead Audit tests your response after hours and records the result. If you want a chatbot built and tuned for you, see our AI chatbot for websites.