The first thing a visitor meets — and the last thing you think about
A chatbot is now the first “person” a customer meets on your website. It pops up before you can pick up the phone, it answers on a Sunday evening, and it never takes a lunch break. That is exactly why it is so easy to switch on — and exactly why it can do damage no other part of the site can.
The mistake most businesses make is not that they use a chatbot. It is that they let it loose on the site before deciding what role it should play. A chatbot that does not know its job starts inventing one. And an invented answer from a tool that speaks for your company is worse than no answer at all.
Where a chatbot quietly costs you customers
The rule is simple: the more a chatbot looks like a promise about your company, the more it costs you when it does not keep it. Here is where that shows up.
When it pretends to be a person. A visitor asks a question and “Anna” replies right away. But Anna never sleeps, never says “I do not know”, and answers in a second at half past two in the morning. The customer figures it out sooner or later — and the moment they realise a machine is playing at being a person, they stop trusting the things on the site that are actually true.
When it makes things up. A chatbot wired only to a general language model does not know your prices, your deadlines, or your terms. But ask it about them and it will not say “I do not know” — it will say something that sounds right. A price it invented, a delivery date that does not exist. The customer walks away with information that is a lie, and you never find out until the complaint arrives.
When it hides the human. A chatbot that answers every question with “let me connect you” but never connects, or one that forces a customer through ten menu options before they find a phone number. Every extra step between a customer and a person is an enquiry you just lost.
When it answers everything the same way. A visitor asks about a specific machine, a specific date, a specific case. The chatbot replies with a generic sentence from your homepage. The customer did not need you to repeat what you do — they needed an answer to their question. A generic answer to a specific question says one thing: this company is not listening.
The test for every chatbot fits in one sentence: if a customer believed what the chatbot said and it was not true, would that hurt us? If the answer is yes, the chatbot must not answer that question.
Where a chatbot genuinely helps
None of this means a chatbot is a bad tool. It is simply the wrong tool for answers that carry a commitment. For answers that save time and promise nothing, it is often better than anything else.
Recurring questions. Opening hours, the address, directions, whether you invoice companies, where to send materials. Questions whose answers you know and that never change. This is the best possible first chatbot — it answers fast, accurately, and from a script you wrote yourself.
Outside working hours. A customer is shopping in the evening or on the weekend while you are asleep. The chatbot answers immediately — and even if the answer is only “send us a message and we will get back to you Monday morning”, the visitor did not go to a competitor, because at least someone responded.
Qualifying enquiries. Before a lead reaches a salesperson, the chatbot asks the essentials: what do you need, by when, what is your budget. The salesperson no longer calls blind but with the context ready. This saves more time than almost anything else — and the chatbot can do it because it only asks the questions you wrote for it.
Routing to the right person. A new enquiry, a technical question, invoicing, a complaint. The chatbot recognises the type of question and sends the customer to the person who can actually help — not one contact for everything, but a specific person for a specific thing.
The pattern is clear: a chatbot is great at what you can describe in advance, and wrong for what has to be decided right now.
The line everything turns on
The most important chatbot decision is not technical. It is a question of trust: tell the visitor they are talking to a machine, and always give them a route to a person.
Companies worry that admitting it is a bot will make customers leave. It is the opposite. Customers do not leave because it is a bot. They leave because the bot pretends to be a person and then lets them down. An honest chatbot that says “I am an automated assistant, here is the phone number for a person” does not lose trust — it builds it.
That line decides everything. Below it, a chatbot sells. Above it, it sends customers away.
A five-minute workflow that avoids the worst mistakes
- Write down what the chatbot may answer. A list of questions and exact answers. Anything not on the list gets “a colleague can tell you” and a handover to a person.
- Switch off invention. Turn off free-form answer generation, or set the chatbot to reply with a fixed phrase for anything outside your material.
- Say it is a bot. In the very first message. No fake names, no “our colleague Anna”.
- Always offer a route to a person. A phone number, email, or form — visible and one click away, not buried under ten menu options.
- Read what the chatbot says. Once a week, skim a few conversations. You will learn what people ask, where the chatbot fails, and which questions your website is missing.
The checklist takes minutes. The damage it prevents is measured in lost enquiries.
A chatbot and a person are not competitors
The strongest websites we see do not choose between a chatbot and a real person. They use both, in the right places. The chatbot handles the repetitive work so a person can focus on what actually has to be solved: complex enquiries, sensitive topics, and customers who are still deciding.
Get that split right and a chatbot saves you hours a week and brings in enquiries even when nobody is at a computer. Get it wrong — let a bot loose that invents prices and hides the phone number — and the best website in the world will not recover the trust you just spent.
If you are not sure where a chatbot would help on your site and where it would hurt, that is exactly the kind of thinking we start every website project with.
Related service: Web design and content strategy