How much does an AI chatbot for a website cost?

A chatbot for your website: what you pay for

An AI chatbot is more than a chat window. To answer helpfully and reliably, it needs a knowledge base built from your content, clear rules for tone and limits, conversation design with handoff to your team, and clean technical integration into the website. The costs of a chatbot are made up of exactly these building blocks.

The knowledge base makes the biggest difference. A bot that is meant to answer questions about opening hours, services, and directions is set up faster than an assistant that has to know extensive product data sheets, price lists, and manuals and answer correctly from them. It matters just as much whether the bot only informs or also books appointments and writes inquiries into the CRM.

• Defining the goal: support, lead qualification, or appointment booking
• Knowledge base from website, FAQ, product data, and documents
• Guardrails against made-up prices and false promises
• Chat widget design matching the brand
• Integration with a form, CRM, calendar, or ticket system
• Testing with real customer questions and evaluation

Rule-based bot or AI chatbot: differences in cost and benefit

Rule-based chatbots work with predefined click paths: the user chooses from answer options, and the bot follows a decision tree. That is inexpensive and predictable but quickly reaches its limits with freely worded questions. AI chatbots, by contrast, understand natural language and respond appropriately – even when customers ask differently than you expected during setup.

So that an AI chatbot doesn't make things up, we generally use retrieval-augmented generation (RAG): the bot first searches for matching passages in your approved knowledge base and formulates its answer only on that basis. If a piece of information is missing, it refers the visitor to your team. This setup costs more than a simple click bot but delivers answers that actually fit your company.

Ongoing costs: AI usage, hosting, and maintenance

Most AI chatbots use a language model through an API. Billing is based on usage, that is, on the amount of text processed, known as tokens. How high these costs are depends on the number and length of conversations, the chosen model, and the amount of content read from the knowledge base for each answer.

On top come hosting for the widget and knowledge base, any licenses for a chat platform, and maintenance: new services, changed opening hours, or prices have to be carried over into the knowledge base. We set up cost limits and reports so that you can see at any time how many conversations are taking place, which questions remain open, and what the AI usage costs.

A free chatbot for your website – where the catch is

Many platforms offer a free chatbot for your website. Usually, tight limits apply to the number of conversations, the widget carries the provider's logo, and connections to CRM or calendar are missing. For a first test of whether your visitors use the chat at all, that is fine – for your company's lasting customer contact, rarely.

Data protection is more critical: with free offerings, it is often unclear where conversation data is stored, whether it is used for training, and whether a data processing agreement is possible. A chatbot that processes your customers' names, email addresses, and requests should therefore run on a contractually sound basis – even if that costs something.

How do I create an AI chatbot – on my own or with an agency?

You can create a simple AI chatbot yourself with chatbot builder tools: upload content, write a greeting, paste a code snippet into the website. Custom assistants can also be created with ChatGPT; however, these run inside ChatGPT. For a chatbot directly on your website, the language model is integrated into your own chat widget via the provider's API.

An agency is worthwhile if the bot is to work reliably, in line with your brand, and in compliance with data protection rules, and be connected to your systems. We handle concept, knowledge base, guardrails, design, integration, and testing – and make sure that the bot hands off to people when it is unsure instead of guessing. That keeps the quality of the answers under control even after launch.

• Define the goal and target audience
• Collect frequent customer questions and approved answers
• Choose a tool or language model
• Write rules for tone and limits
• Test with real questions and refine
• Integrate it and adapt the privacy policy

Data protection and transparency for an AI chatbot

A website chatbot processes personal data as soon as visitors enter names, email addresses, or personal requests. A legal basis, a note in the privacy policy, a data processing agreement with the providers used, and defined deletion periods for chat logs are therefore required. If data is transferred to countries outside the EU, this must be additionally safeguarded, for example through standard contractual clauses or an adequacy decision.

The EU AI Act also provides for transparency obligations: users should be able to recognize that they are communicating with an AI. A clear notice in the chat window is therefore good practice anyway and strengthens trust. These notes do not replace legal advice; coordinate the use with your data protection officer before the chatbot goes live.

How to recognize a chatbot that pays off

A chatbot pays off when it reliably answers recurring questions, hands qualified inquiries to your team, and leads visitors to a next step – such as booking an appointment or the quote form. So don't just measure the number of conversations but also how many of them were resolved and how many concrete inquiries resulted.

Just as important is what the bot does not do: it doesn't make up prices, make binding commitments, or give legal or health advice. We define these limits together and check them before launch with typical and deliberately critical test questions. That way the chatbot protects trust in your company instead of endangering it with wrong answers.