AI chatbot integration means adding a chat window to your website in which an AI model answers visitors from your own business information: services, hours, shipping, warranties, frequently asked questions. You can integrate an AI chatbot into a website in three ways: through a chat platform that already has an AI agent and goes on your site with a snippet of code, through an AI agent builder you train on your own pages and documents, or through a custom integration with a model’s API from OpenAI, Anthropic or Google, connected to your site, your CRM and your messaging apps. Whichever you choose, three things decide the result: the information it answers from, when it hands the conversation to a person, and where the contact details left in the chat end up.

Many of your visitors already talk to AI: about half of US adults say they use AI chatbots such as ChatGPT, Gemini or Copilot, and 44% use ChatGPT itself (Pew Research Center, survey of February 2026). Among US businesses, 19.8% said they had used AI in the past two weeks, in the Census Bureau’s survey period ending May 3, 2026.

Below: the three ways in detail, the steps, the 2026 rules and how to measure the result.

What an AI chatbot on your website can do

  • Answers the questions that keep coming back: hours, address, shipping, warranty, how to book an appointment.
  • Finds out what the visitor is looking for and recommends the right service or product.
  • Takes a name and a phone number or email and sends them to your CRM, with a summary of the conversation.
  • Books a meeting in your team’s calendar.
  • Answers in the language the visitor writes in, such as English or Spanish.
  • Takes conversations after hours, so your team finds them in the morning.

What it must not do: invent prices, deadlines or promises it cannot find in your information, ask for sensitive data without a reason, or keep someone in the chat when they want to talk to a person on your team.

Three ways to integrate an AI chatbot into your website

1. A chat platform with a built-in AI agent

Customer service platforms such as Intercom or Zendesk already have an AI agent that answers in their chat from your help center and policies. You add it to your site with a snippet of code or a plugin. It starts quickly, and the handover to a human agent is already built in. In return, you pay the platform’s subscription (some also charge a fee for each conversation the AI resolves), and the conversations stay with the vendor.

If you sell mostly through messaging apps, the assistant has to answer there too, and CRMs built for that have their own AI agents: Kommo’s, for example, answers on WhatsApp, Instagram and TikTok and moves leads through your sales pipeline.

2. An AI agent builder trained on your documents

Services such as Chatbase or Voiceflow let you add your website pages, documents and FAQs, write the agent’s instructions and put the chat on your site by pasting a snippet, without programming. They suit an assistant that answers questions. Connecting it to your CRM, calendar or inventory usually takes extra integrations.

3. A custom integration through an API

A developer connects your site directly to an AI model’s API and to your systems: CRM, catalog, bookings. The chatbot first looks up the answer in your documents and only then writes it, and you set the rules: what it may say, when it hands over, what data it asks for. It takes longer to launch, but the model is billed by the amount of text it processes, and you do not depend on a chat platform. The same assistant can also answer on WhatsApp or Facebook Messenger.

AI chatbot integration, step by step

  1. Pick one job to start with: answering frequent questions, qualifying leads or booking appointments. A chatbot that does one thing well brings more than one that tries to do everything.
  2. Gather the information it answers from: service pages, frequently asked questions, shipping and return terms, opening hours, in every language you serve. Keep it current: the chatbot repeats what it finds, outdated information included.
  3. Write the rules: your brand’s tone, the topics it does not discuss, what it says when it does not know (“I don’t have that information. Let me connect you with a colleague.”) and when it asks for contact details.
  4. Set up the handover to a person: during business hours, to someone in the chat or on the phone; after hours, the chatbot takes the contact details and says when the person will hear back. Decide who answers, and how fast.
  5. Send leads to your CRM: a visitor becomes a lead only when they leave a phone number or an email. The chatbot sends the details and a summary of the conversation to your CRM and alerts your team.
  6. Test it with real questions, including ones it must not answer and attempts to make it promise a discount.
  7. Launch, then read the conversations every week for the first month. The questions it could not answer show you what is missing from your site.

Looking for a Results-Driven Marketing Agency?

Get measurable results within your budget.
Partner with Us
 

What the rules require in 2026

  • Say that an AI is answering. In the EU, Article 50 of the AI Act applies from August 2, 2026: people must be told from their first interaction that they are talking to an AI system, unless that is obvious. The duty falls on the provider, even one based outside the EU, when its chatbot answers people in the EU: the platform, if you use one (check that it does this), or you, if you have your own chatbot built and run it under your name.
  • In California, never let a bot pass for a person. Since July 1, 2019, it has been unlawful to use a bot online with the intent to mislead people in California about its artificial identity, in order to encourage a purchase or sale. A business that discloses the bot, clearly and conspicuously, is not liable under that law.
  • Wherever you operate, say it in the chatbot’s first message: “I’m an AI assistant. I can answer your questions or connect you with our team.” It costs nothing and avoids misunderstandings.
  • Protect the data in conversations. Chats hold names, phone numbers and sometimes other personal information. Mention the chat in your privacy policy and link to it in the chat window, sign a data processing agreement with the vendor, and find out where conversations are stored and for how long. If you offer goods or services to people in the EU, the GDPR applies even if you are based outside it, and it requires that agreement. California’s CCPA covers for-profit businesses doing business in the state that meet one of its thresholds, such as buying, selling or sharing the personal information of 100,000 or more California residents or households; those residents have the right to be told what you collect and how you may use it, at or before collection.

How to measure whether it works

  • how many conversations start, and on which pages;
  • how many end with an answer, how many with a handover to a person and how many with a lead;
  • which questions go unanswered;
  • in Google Analytics 4: opening the chat and a lead left in the chat as separate events, with the lead marked as a key event.

One more thing, for your site’s speed: load the chat window only on the visitor’s first scroll, tap or key press, not with the page. We do the same with other third-party scripts that the first view does not need, because on phones they delay the page.

We build AI assistants trained on your documents and connected to the tools you already use through our AI integration service.

Common mistakes

  • The chatbot answers from the model’s general knowledge, not from your business information, and invents prices or terms.
  • There is no “Talk to a person” button.
  • It asks for a phone number before answering anything, and the visitor closes the window.
  • It answers in English to someone who writes in Spanish, or the other way around.
  • The window opens by itself and covers the page on a phone.
  • Contact details stay in the chat platform and never reach anyone on your team.

Frequently asked questions

How much does an AI chatbot on a website cost?

The cost has three parts: the platform’s subscription or the model’s usage, billed by the amount of text; with some platforms, a fee for each conversation the AI resolves; and the setup work: preparing the information, connecting your CRM and testing. Platforms publish their prices.

Can it answer in more than one language?

Yes. Give it your information in each language, especially product names, technical terms and conditions, and test the answers in each language separately.

Does a chatbot replace a live agent?

No. It takes the repeated questions and the conversations after hours, and your people stay for sales, complaints and unusual cases.

Do I have to tell visitors they are talking to an AI?

In the EU, yes, from August 2, 2026, unless it is obvious, and in California a bot that sells must not pass for a person. Wherever you operate, say it in the first message.

Sources