Adding a Chatbot to Your Website: Real Value and Options
A chatbot on a website automatically answers visitors' frequent questions, usually through a chat window at the bottom of the screen, available outside business hours. Common SaaS solutions cost between 20 and 100 euros per month; a custom chatbot with complex business logic requires development running into thousands of euros. It's only worthwhile once there's a real volume of repetitive questions.
The real problem: chatbots are often added before they're needed
The most common mistake is installing a chatbot because "it's modern" or because competitors have one, without having identified the actual volume of questions it would handle. The result: a chatbot that answers poorly because it wasn't fed enough relevant content, or a correctly configured chatbot visited by too few people to justify its monthly cost.
Before adding a chatbot, it's worth answering a simple question: what are the five to ten questions your visitors ask most often, by phone or email? If that list exists and repeats, a chatbot makes sense. If requests are different every time and need a real personalized exchange, a chatbot won't add much.
The three families of chatbot
Scripted chatbot (decision tree)
Works with buttons and closed questions: "Are you looking to: book an appointment / see our pricing / something else". Predictable and easy to maintain, but limited to paths planned in advance. Works well for directing a visitor to the right page or contact.
Chatbot with a knowledge base (automated FAQ)
Answers freely worded questions by matching them against a pre-prepared set of questions and answers. More flexible than a decision tree, but still limited to the topics covered in the base.
Chatbot based on generative AI
Understands naturally worded questions and generates a constructed answer, not just one picked from a preset list. More convincing for the user, but requires rigorous framing: without clearly defined limits, this type of tool can produce inaccurate answers, invent information, or stray outside its intended scope (wrong pricing, unkept promises).
Comparing the options
| Solution | Typical cost | Setup | Good for |
|---|---|---|---|
| Chat widget with offline message (no AI) | €0 to €30/month | A few hours | Redirecting to email/phone when no one is available |
| Scripted SaaS chatbot (Tidio, Crisp) | €20 to €60/month | A few days | Guiding the visitor, answering the most frequent questions |
| Generative AI SaaS chatbot | €40 to €100/month depending on volume | 1 to 2 weeks of configuration | Sites with varied questions, need for natural answers |
| Custom chatbot with business logic | €2,000 to €10,000 and up | 1 to 3 months | Integration with a proprietary system, very specific use case |
What separates a useful chatbot from a gimmick
- A clear scope: the chatbot should know how to say "I don't know, here's how to reach a person" rather than inventing an approximate answer.
- Real content behind it: a generative AI chatbot works better when it draws on your actual content (website pages, existing FAQ) rather than generic knowledge.
- A gateway to a human: for any sensitive or complex question, a clear button or link to a human contact avoids frustration.
- Regular monitoring: reviewing the questions asked to the chatbot often reveals needs poorly covered by the rest of the site, to fix directly on the relevant pages.
When a chatbot isn't the right answer
- For a low-traffic site, the monthly investment often exceeds the real value it brings: a visible phone number or a well-designed contact form is enough.
- For questions that require real professional judgment (legal advice, medical diagnosis, complex financial situations), a chatbot should never substitute for human exchange, and could even create liability if it gives an inaccurate answer perceived as binding advice.
- For a sector where trust is built through direct contact (tradespeople, healthcare professionals), a chatbot can create an impression of distance rather than service, depending on the target audience.
Setting up a chatbot step by step
- List the five to ten most frequently asked questions from your visitors or clients, based on your actual history (emails, calls, forms).
- Choose the right level of sophistication: scripted for simple guidance, knowledge base or generative AI for more varied questions.
- Write the answers yourself or with a provider's help, drawing on your existing content (website pages, FAQ) rather than generic answers.
- Clearly define the limits: which questions should always redirect to a human contact rather than receive an automated answer.
- Test the chatbot with real questions, including awkwardly worded ones, to check it doesn't answer off-topic or invent information.
- Monitor conversations during the first few weeks to adjust insufficient answers and spot unanticipated questions.
The risk of misplaced trust
A well-designed chatbot inspires confidence, which can become a trap if it gives inaccurate information with the same assurance as accurate information. Unlike a human who can express doubt ("I'm not sure, let me check"), a poorly framed chatbot answers with the same fluency whether it's right or wrong. This is a particular point of attention for generative AI chatbots, which must be explicitly limited to a verified scope of knowledge rather than left free to generate an answer on any subject.
What to remember
- A chatbot only makes sense once a real volume of repetitive questions has been identified, not by default.
- Three levels exist: scripted (simple and predictable), knowledge base (flexible but limited), generative AI (natural but needs careful framing).
- Common SaaS solutions cost between 20 and 100 euros per month, plenty for most professional sites.
- A well-designed chatbot knows to redirect to a human rather than invent an answer when it doesn't know.
- For sensitive topics (legal, health, finance), the chatbot should point toward a human exchange rather than answer in place of a professional.
Frequently asked questions
Does a chatbot replace real customer service? No, not entirely. A well-designed chatbot handles repetitive questions and qualifies more complex requests before passing them on to a person. For personalized advice, it should redirect to a human contact.
Is an AI chatbot expensive to set up? Current SaaS solutions generally cost between 20 and 100 euros per month depending on conversation volume. A fully custom chatbot with complex business logic runs into thousands of euros in development.
Does a small showcase website need a chatbot? Not necessarily. For a site with few visitors or rarely repeated questions, a simple contact form or a visible phone number is largely enough.
What's the difference between a scripted chatbot and a generative AI chatbot? A scripted chatbot follows pre-set decision trees: predictable but rigid. A generative AI chatbot understands freely worded questions and answers more naturally, but requires careful framing to avoid inaccurate answers.
In summary
A chatbot is worth its monthly cost only if a real volume of repetitive questions justifies it: identify that before installing anything. If you're looking to improve your first contact channel before adding a chatbot, the VeryAppi website plan already includes a contact form and an FAQ structure designed to reduce repetitive questions.
Frequently asked questions
›Does a chatbot replace real customer service?
No, not entirely. A well-designed chatbot handles repetitive questions (hours, pricing, availability) and qualifies more complex requests before passing them on to a person. For personalized advice or a sensitive situation, it should redirect to a human contact rather than trying to handle everything alone.
›Is an AI chatbot expensive to set up?
Current SaaS solutions (Crisp, Intercom, Tidio, or tools built on language models) generally cost between 20 and 100 euros per month depending on conversation volume. A fully custom chatbot with complex business logic, on the other hand, runs into thousands of euros in development.
›Does a small showcase website need a chatbot?
Not necessarily. For a site with few visitors or rarely repeated questions, a simple contact form or a visible phone number is largely enough. A chatbot becomes worthwhile once there's a real volume of repeated questions that justifies automation.
›What's the difference between a scripted chatbot and a generative AI chatbot?
A scripted chatbot follows pre-set decision trees (buttons, closed questions): predictable but rigid. A chatbot built on generative AI understands freely worded questions and answers more naturally, but requires careful framing to avoid inaccurate or off-topic responses.