A chatbot is software that holds a written conversation with people and replies without a person typing. In an online store it usually appears as a chat window in the corner of the website or as a bot on the business's WhatsApp number, answering questions like how much shipping costs, how returns work and which size to pick. The word covers very different tools, from a numbered menu that reacts only to set choices to an AI chatbot that understands free text and writes its own reply. The difference decides whether a customer gets a useful answer or gets stuck. Below: what a chatbot is, the main types, how an AI chatbot answers, and the mistakes to avoid.
What is a chatbot
A chatbot is an automated assistant in text: the customer sends a message and the software replies at once. For a business, the key question is whether it answers only from the business's own information.
- Where it appears: on a website, on WhatsApp or in an app.
- What it answers: shipping, returns, sizing and other common questions.
- What it is not: a human rep. With live chat, a person on your team replies.
Types of chatbot
- Rules chatbot. Works from a menu, buttons or keywords. If the customer types "shipping", it sends the shipping answer. Predictable and cheap, but it breaks when a customer writes a sentence nobody planned for.
- AI chatbot. Uses a language model to understand the message and write a reply. Copes with typos and two questions in one message. The risk is an answer that sounds confident and is wrong.
- Hybrid chatbot. Combines buttons for the most common questions with a language model for everything else.
How an AI chatbot answers
An AI chatbot built for a business usually works in three steps:
- It searches the business's knowledge base for passages that match the question, such as the shipping policy page.
- It hands those passages to the language model.
- The model writes a reply based on them, in the language the customer wrote in.
This approach is often called RAG, short for retrieval-augmented generation: writing that rests on search.
Example: a size exchange
- A customer writes: "I ordered a medium, can I switch to a large?"
- The chatbot finds the exchange policy.
- It replies that sizes can be exchanged within the number of days stated there, with a link to the page.
- If the policy is not written anywhere, a good chatbot says it does not have that information and points to the team, instead of inventing a number of days.
The customer gets an instant answer, and the business does not commit to something it never wrote.
Chatbot vs AI agent
The terms overlap, and there is no single agreed definition. "AI agent" usually describes a chatbot that can also act rather than only answer:
- Hand a conversation to the team.
- Follow rules and run in different modes.
- Use other tools.
When a vendor shows you an "agent", ask exactly what it does beyond replying.
Common mistakes
- A chatbot without current knowledge. It will answer from an old shipping page, and the promise binds the business.
- No route to a person. A customer stuck with a bot and no way out leaves, or writes a review.
- Hiding that it is a bot. It is better to tell customers they are talking to an automated assistant.
- Letting it decide about money. Refunds, compensation and discounts stay with the team, even when the bot answers everything else.
- Expecting it to know live data. A chatbot not connected to orders and stock cannot know where a specific parcel is or what is left on the shelf.
Chatbots in StoreChart
StoreChart has two separate tools of this kind:
- Bubble, an AI chatbot for your website: answers from a knowledge base built by scanning your site, recommends products from the scanned catalog, and points visitors to your store's WhatsApp or phone when it has no answer. It does not check stock or hand off to a live rep.
- The WhatsApp AI agent: answers customers on your business number through the WhatsApp Business API, and can hand a conversation to your team.
For practical advice on choosing and installing one, see the guide to choosing a chatbot for your website.
Frequently asked questions
Chatbot vs live chat: what is the difference?
With a chatbot, software replies; with live chat, a person on your team does. A chatbot is available at any hour and answers instantly, but only knows what it was given. Live chat handles complex questions, but depends on someone being at the screen.
Can an AI chatbot make up answers?
Yes, if it is not restricted. A language model that cannot find an answer tends to write something that sounds right. That is why a business chatbot should answer only from the business's knowledge base and say plainly when the information is not there.
Do I need to code to run a chatbot?
No. Most business tools install on your site with one line of code pasted into the theme, and the knowledge comes from a scan of your site and text you write in the dashboard. The real work is keeping that content correct and current.