AI live chat on your website: how to set it up

AI live chat is a chat window on your website in which an AI agent answers customers in real time, from your products, policies and orders, at any hour. You set it up in seven steps: decide what the AI answers, give it your data, set its instructions, install the chat widget, test it on real questions, set up the hand-over to a person and read the conversations every week. It matters most outside office hours: about 42% of conversations on Chaterimo started between 18:00 and 8:00 Prague time (February–May 2026, Chaterimo platform statistics).

Chaterimo is our product. It is an AI agent, not a live chat with people; we say below where it fits and where a live chat tool is the better choice.

What is AI live chat?

The term is used for different things, so it helps to name them first.

Live chat

A person from your team answers the customer in the chat window, in real time. It works only while someone is online.

AI chatbot (AI agent)

A language model (LLM) answers in the chat window from your product catalogue, your policies and, when your store is connected, your orders. It answers at any hour and many customers at once. Older chatbots followed a scripted flow with buttons; an AI agent understands questions in the customer's own words.

AI live chat

Usually an AI agent in the chat window on your website, with a way to reach a person. Some vendors mean live chat software in which the AI answers first and an operator can take over the same chat window. Others hand the conversation over later, as a ticket or an e-mail.

When you compare tools, ask the vendor which of these it means. The answer decides who on your team has to be online, and when.

Live chat, AI chatbot or both: which fits your store?

Choose live chat when your team can be online in the hours your customers write and your questions need a person: custom orders, complaints, B2B prices, negotiation. A live chat window with nobody online is worse than none, because the customer waits for an answer that does not come.

Choose an AI chatbot when most questions repeat (delivery, returns, stock, which product fits) or when nobody can watch the chat during the day. It also suits a large catalogue: the AI searches thousands of products for the one that matches the question.

Combine both when you have a team, but not around the clock. The AI answers first and at night, and a person takes over what needs a decision. Decide whether the person takes over in the same chat window (live chat software with an AI agent) or later by e-mail or phone (a ticket). Our comparison of live chat apps for Shopify covers tools of the first kind.

How fast do live chat and an AI chatbot answer?

Text, the company behind LiveChat, reports an average first response time of 35 seconds across its customers, and about 20 seconds more in retail (data last updated in 2024)¹. That holds only while an operator is online. When nobody is, the customer leaves a message and gets the reply hours later, or leaves the store.

An AI chatbot starts answering within seconds, at any hour. In our blind test of 12 language models on 45 real e-commerce support questions (round 4, 30 August 2026), 8 models answered in under 5 seconds on average and the slowest needed 49.7 seconds. The details are in the Chaterimo CX-LLM Benchmark. Speed differs by model, so check it before you choose one; the guide to choosing an LLM for customer support explains the trade-off between speed, accuracy and price.

When do customers write to an online store?

Across Chaterimo's paying stores, about 42% of 452,568 conversations started between 18:00 and 8:00 Prague time, on any day of the week (February–May 2026, Chaterimo platform statistics). The methodology is in our report on the state of AI customer support in e-commerce.

A team that works office hours does not see four in ten of these questions until the next morning. If you staff live chat only during the day, the evening and night questions need another answer: an AI agent, or at least an honest offline message that says when you will reply.

What does live chat cost compared with an AI chatbot?

Live chat software is usually priced per operator or seat, sometimes per conversation or ticket. AI agents inside live chat tools mostly cost extra, per AI conversation or per resolution. Our comparison of live chat apps for Shopify lists the prices of twelve apps, checked on 20 October 2026.

The larger cost of live chat is people. Someone has to be online for every hour you show the chat as available, including evenings if you want to cover them. Count the hours you would staff, not only the licences.

An AI chatbot costs a monthly plan and the use of the language model. Some vendors include a number of AI messages in the plan, others charge per AI conversation or resolution. With your own API key (BYOK: you use your own key and pay the model provider directly), you pay the provider for what the chatbot uses. Chaterimo has plans of both kinds; see the pricing page.

To compare, take last month's conversations. What would that volume cost with operators online in the hours your customers write, and what with an AI agent plus a person for the rest?

When does an AI chatbot need a person?

An AI agent answers from the data you give it. It needs a person when:

• the answer is not in its data, for example delivery to a country you do not list;
• the customer wants a decision: a refund outside your policy, a discount, an exception;
• the customer complains and wants to be heard by a person;
• the request needs an action the AI cannot take, such as changing an order.

A fluent answer is not a correct one. In our benchmark, accuracy was the weakest criterion for all 12 models: the best scored 55.4 out of 100 on accuracy, while every model scored between 81 and 90 on tone (round 4, 30 August 2026). So test the answers before customers see them, and keep the way to a person open.

Tell the AI when to hand over, and tell the customer what happens next: who replies, through which channel and by when.

How do you set up AI live chat on your website?

These seven steps apply to any AI chat tool. The clicks in Chaterimo are in our guide to your first AI chatbot in 4 steps.

Where does Chaterimo fit?

Chaterimo is an AI agent for customer support of online stores. It answers around the clock in the chat window on your store and in Facebook Messenger, recommends products from your catalogue and, with a connected store, tells customers the status of their order once they give the order number and the e-mail or phone number used on it. It works with models from OpenAI, Anthropic, Google and xAI, and you can use your own API key (BYOK).

Chaterimo has no live chat: nobody from your team takes over the chat window. A conversation reaches a person as a ticket with the conversation and an e-mail, and your team answers by e-mail or phone. Chaterimo fits stores where most questions are about products, delivery and orders, and where nobody can watch the chat all day. If your team wants to talk to customers live, a live chat tool fits better; compare the live chat apps for Shopify or read Chaterimo vs Tidio.

Summary

AI live chat puts an AI agent into the chat window on your website, so customers get an answer within seconds at any hour. Live chat with people answers only while someone is online, and about 42% of conversations on Chaterimo start between 18:00 and 8:00 Prague time. Choose live chat for questions that need a person, an AI chatbot for repeated questions and the hours nobody is online, and both when you have a team but not around the clock. Then give the AI your data, install the chat widget, test it on real questions and decide how a conversation reaches a person; the guide to your first AI chatbot shows these steps in Chaterimo. What an AI chatbot can do for an online store beyond the chat window is in the overview of AI chatbots for e-commerce.

TopicsAI customer service