Can an AI chatbot on your website give visitors a price? What works and what doesn’t

Someone lands on your website at nine in the evening, reads two pages and opens the chat. The first thing they type is “how much for a rewire?” or “roughly what would a five-page site cost?”. They want to know whether you are in their budget before they give you their name. This guide sets out whether a chatbot should answer, the three ways it can, and the risks that come with each.

Why visitors ask for a price in chat

A price is the fastest way to qualify a supplier. A visitor comparing three firms will ask all three the same question and keep talking to the ones that answer. A chat window invites the question because it feels like talking to a person, and because the alternative, a form, means waiting a day for a reply.

The visitor is usually not asking for a contract price. They are asking whether you are a £500 firm or a £5,000 firm. The mistake most businesses make is to treat that question as if it needed a firm quote, and either refuse to answer or answer with a number they later regret.

Ballpark, estimate and quote are different things

The words get used loosely, and chat is where loose words cause trouble.

A quote is a price for defined work. Once accepted, you are expected to do that work for that price unless the scope changes. An estimate is your best judgement of what the work will cost, and it may change as the job reveals itself. A ballpark is a rough, non-binding range offered to help someone decide whether to continue the conversation.

Only one of the three belongs in a chat window before a human has looked at the job, and that is the ballpark, clearly labelled as such. A quote needs defined scope, assumptions and exclusions, and those come after questions, not before. For the full difference and how to write each, see how to quote a job.

The three ways a chatbot can answer

Never. The bot explains that prices depend on the job, asks its questions, takes contact details and promises a quote. This is safe, and it still beats a form, because the visitor leaves with their enquiry fully scoped rather than half described.

A range from your rates. The bot asks enough to know what kind of job it is, then gives a non-binding ballpark drawn from your actual rates and past quotes, with wording that makes clear the firm price follows. This answers the visitor’s real question, “am I in budget?”, without committing you. It works when your pricing has a recognisable shape, for example a day rate, a price per room or a price per bedroom.

A firm itemised quote after questions. The bot runs the whole discovery conversation, then produces a full quote with line items, quantities, rates, assumptions, exclusions and VAT. Whether that quote is shown in chat or sent as a document for you to approve first is the important choice. ChatQuote, for example, drafts the itemised quote and branded PDF proposal from the conversation but holds it for your “Approve & send” unless you switch on automatic sending. The visitor gets a fast, complete answer and you still get the final say.

The risks, and where they come from

A wrong firm price in chat. If a bot states a firm figure and the job turns out to be bigger, you either absorb the difference or start the relationship with an argument. The fix is labelling. A range offered as a ballpark is a help. The same range offered as a quote is a liability.

Scripted bots. Many website chatbots are decision trees. They can only ask the questions someone wrote down in advance and cannot cope with “it’s a Victorian terrace but the back half was rebuilt in the nineties”. They give useless prices because they cannot hear what makes this job different.

Generic ChatGPT. A general assistant can lay out a quotation neatly once you paste in the scope and the prices. On your website it knows neither. Without your rates, your past quotes and the questions that drive your pricing, it will guess, and a confident guess is the most dangerous kind of answer.

Over-asking. A bot that fires twenty questions before offering anything loses the visitor. The questions should be the ones that actually change the price, asked a few at a time.

What the bot needs to know

To give a price that is useful and safe, a chatbot needs three things.

Your rates. Day rates, hourly rates, per-unit prices, fixed fees or monthly retainers, whichever you use, plus your VAT position and payment terms.

Your past quotes. A handful of real quotes teach a system how you actually price, which is often different from how you think you price. ChatQuote learns from past quotes uploaded as PDF, text, Markdown or CSV, and from your website, then shows you what it learned so you can correct it in plain English, for example “our day rate is now £520”.

The questions that change the price. For an electrician, how many circuits and whether the consumer unit is being replaced. For a removals firm, how many bedrooms and whether packing is included. For a decorator, which rooms and what state the walls are in. For a web agency, how many pages and which integrations. The pages for electricians, removals, painters and decorators and web agencies show the kinds of questions that matter in each line of work.

What happens after the chat

A price in chat is only half the job. The useful part is what the business receives.

At minimum, every chat should become a lead with a transcript, including the ones the visitor abandoned halfway. A visitor who gave their email and three answers before drifting off is still a warm enquiry, and the transcript saves you asking the same questions twice.

Better still, the conversation should become a draft quote. With ChatQuote, each finished chat leaves an itemised draft and a branded proposal waiting on the lead’s page for review. You check the lines, change what needs changing and send. Unfinished chats still become leads with a transcript, so nothing is lost.

When not to show prices in chat

There are good reasons to keep prices out of the chat entirely.

Your pricing is confidential or negotiated per client, as with many marketing agencies and consultancies. Your jobs vary so much that any range would be meaningless. Your competitors are likely to use your chat to benchmark you. Your sales process depends on a site visit, and a number given beforehand would anchor the customer before you have seen the job.

In all of these cases the right setting is the first option above: no prices in chat, full questions, contact details captured, quote to follow. ChatQuote treats this as a switch, ballpark on or no prices, and the chat does the scoping either way. For how this compares with a form or a price calculator, see a website chatbot that gives a price.

Frequently asked questions

Is a price given by a chatbot legally binding?

A price presented clearly as a non-binding ballpark or estimate is not an offer to do the work at that figure. A price presented as a quote for defined work may be treated as one once the customer accepts it. Label ballparks as ballparks, keep firm quotes for the document you review, and if in doubt take advice on your own terms of business.

Can the chatbot give a price without my rates?

It can produce a number, but not a useful one. A chatbot that does not know your rates is guessing from general knowledge, and it will be wrong in both directions. Give it your rates and a few past quotes before you let it mention money.

Will visitors see my other customers’ prices?

They should not, and with ChatQuote they do not. Past quotes teach the assistant how you price and are never given to the assistant that talks to prospects. With ballparks switched off, even your rates are withheld from it.