There's no single answer to “how much does a website chatbot cost” – the price depends on what the chatbot should do, what content it answers from and what it needs to connect to. This guide explains what a modern AI chatbot can handle, what the price is made of, how to choose between an off-the-shelf tool and a custom build, which obligations not to forget, and how to work out whether a chatbot pays off for you.
What is an AI chatbot for your website
| Scripted chatbot | AI chatbot | |
|---|---|---|
| How it works | Buttons and pre-built branches | Understands questions in everyday language |
| Unexpected question | Can't answer | Looks the answer up in your content |
| Maintenance | Every new branch has to be built | Just update the content |
| Languages | Each language separately | Handles several languages at once |
An AI chatbot is a chat window on your website or online store where visitors type questions in everyday language and get answers instantly – evenings and weekends too. Unlike older chatbots, it doesn't follow pre-built click-through scripts; it uses a large language model (LLM) that understands a question even when it's asked differently than you expected.
A good business chatbot doesn't answer “from memory”, though. Before answering, it looks up information in your own content – FAQs, price list, terms and conditions or product catalogue – and talks only about your company. This approach is called RAG (retrieval-augmented generation).
What a website chatbot can do
What a chatbot shouldn't do is decide high-risk matters (complaints with financial impact, medical or legal advice) without human review. There, it should direct the customer to a person.
- answer questions about prices, availability, shipping, opening hours or terms
- help customers choose a product or service based on their needs
- report order or booking status, if connected to your system
- capture enquiries and contacts outside business hours and send them to your CRM or email
- book an appointment or link to your booking calendar
- hand the conversation to a person, with its full history, when it can't help
- internal use: answer employees from company policies and documents
What an AI chatbot's price is made of
One-off setup
Deciding what the chatbot should answer, preparing and cleaning up the content (FAQs, price lists, terms), setting its behaviour and tone, testing it on real questions and embedding it in your website or store. For a custom chatbot, this is the largest part of the price.
Integrations
Connections to your online store, CRM, booking or stock system. With them, the chatbot doesn't just answer – it can check an order status, record an enquiry or offer a free slot.
Running costs
Fees for using the AI model, which grow with the number of conversations, plus hosting. With off-the-shelf tools, these are included in the monthly subscription. At the usual enquiry volume of a small or medium-sized business, they are typically the smaller part of the total cost.
Maintenance and improvement
Updating the content when prices or services change, and regularly improving answers based on real conversations. A chatbot nobody looks after goes out of date.
Three levels of solution and what they cost
| What you get | How you pay | Who it's for | |
|---|---|---|---|
| Off-the-shelf tool | A chatbot you set up yourself on your FAQs | Monthly subscription, usually tens to hundreds of euros depending on conversations | A small website with a few recurring questions |
| Off-the-shelf tool, set up for you | A tool someone configures, fills with your content and tests | Subscription + one-off setup | A company that doesn't want to handle the setup itself |
| Custom chatbot | A solution designed around your processes, with integrations and human handover | One-off setup + running costs by usage | Online stores and companies where the chatbot must work with your data and systems |
The cheapest option isn't always the cheapest overall. If an off-the-shelf tool can't answer half the questions, customers still end up with your team – just more frustrated.
What raises the price and what lowers it
- Scope and quality of your content: the better prepared it is, the less setup work.
- Integrations: connecting to a CRM, online store or booking system adds work, but also value.
- Languages: every additional language needs testing.
- Handover to a person: routing the conversation to your team with its full context.
- Number of channels: website, online store, social media or internal use for employees.
- Security and data requirements: where data is processed and how long it's kept.
Off-the-shelf tool or a custom chatbot?
| Off-the-shelf chatbot tool | Custom chatbot | |
|---|---|---|
| Launch | Fast, you set it up yourself | Designed and set up for your company |
| Price | Monthly fee | One-off setup + running costs |
| Customisation | Within the tool's limits | Built around your processes |
| Connection to your systems | Supported integrations only | With whatever you use |
| Ownership | Tied to the tool's vendor | The solution and content stay yours |
An off-the-shelf tool is a good start if you only need to answer a few frequent questions. A custom chatbot makes sense when it has to work with your data, connect to your systems or handle more complex situations.
How to choose a chatbot: 10 questions for your supplier
- What content will the chatbot answer from, and who prepares it?
- What does it do when it doesn't know the answer – make one up, or hand over to a person?
- How will it be tested on your customers' real questions before launch?
- Which of your systems can it connect to (online store, CRM, bookings)?
- Which languages does it answer in, and how well?
- Where are conversations and personal data processed and stored?
- How does the chatbot tell users they are talking to AI?
- Will you see conversation history and statistics?
- Who updates the content, and how often, when prices or services change?
- What is the monthly price made of, and how does it grow with the number of conversations?
Obligations: the AI Act and GDPR
From 2 August 2026, Article 50 of the EU regulation on artificial intelligence (the AI Act) applies: people must know they are interacting with an AI system unless it's obvious from the circumstances. In practice, a clear label such as “I'm an AI assistant” right in the chat window is enough.
If the chatbot processes personal data – a name, email, phone number or the content of an enquiry – GDPR applies. Mention the conversations in your privacy policy, sign a data processing agreement with your supplier and keep only what you really need.
How to launch a chatbot step by step
- Write down the 20–50 questions customers ask most often, with the correct answers.
- Add supporting content: price list, terms, shipping, products or services.
- Decide when the chatbot should hand the conversation to a person, and to whom.
- Test the chatbot on real questions, including the awkward ones.
- Launch it on part of the website first, or with your team watching.
- For the first few weeks, read the conversations and fill gaps in the content wherever the chatbot hesitated.
How to measure whether the chatbot works
- the share of conversations the chatbot resolved without a person
- the number of handovers to a person, and why
- enquiries and contacts captured outside business hours
- the time your team saved on recurring questions
- customer feedback after the conversation
How to work out whether a chatbot pays off
Work out how much time your team spends each week answering recurring questions and multiply it by the hourly cost. Add the enquiries that arrive in the evenings and at weekends and go unanswered today.
Example: if the team spends 10 hours a week answering the same questions and the chatbot handles half of them, you save roughly 20 hours a month – and customers get an answer instantly.
What an AI chatbot costs at MacZaj
We build chatbots to measure, so the price is custom too. After the initial consultation we send a detailed quote within 24 hours. A 30% deposit is paid before we start, the rest on completion. Prices exclude VAT.