FAQ Chatbot on Your Website: What It Should Answer and What It Should Hand Off

An FAQ chatbot fails in two opposite ways. It answers too little, sending every visitor to a contact form. Or it answers too much, guessing at questions it should never touch. Setting the scope on paper before you configure anything prevents both.

The phone version of this problem is covered in training your AI receptionist on your FAQs. A website bot has a different advantage. It can link to pages, show the booking flow inline, and hand a visitor a form at the exact moment a question turns personal.

Answer these in chat, every time

Opening hours, location and parking, which services you offer, how booking works, cancellation policy in plain terms, what to bring or prepare, and how long a typical appointment takes. These questions have stable answers that live on your site already. The bot should give the answer in chat rather than pasting a link alone, because a visitor who has to leave the chat to read a page often does not come back.

Keep each answer short. Two or three sentences, then an offer to book or to keep asking. Long pasted paragraphs read like a brochure and visitors skim past the part they needed.

Route these to a person or a form

Anything tied to one customer record belongs elsewhere. Billing disputes, refund requests, complaints, medical or legal specifics, and questions about an existing order or case. The bot can acknowledge the topic, explain that a team member handles it, and collect a name and number with a short description. Guessing at account specific answers creates real harm when the guess is wrong, and a confident wrong answer is worse than a clear handoff.

Pricing sits in the middle. Published prices can be quoted. Estimates that depend on the job should be framed as a range only if you publish one, or routed to a quote request.

Find the gaps in your chat logs

Once a month, read the questions the bot could not answer. Group them. Some are new FAQs worth adding, such as a service you launched recently. Some are wording problems, where visitors ask a known question in a way the bot does not recognize. Some reveal pages your site is missing entirely. The log is the cheapest research you will ever do on what customers want to know, because they are telling you in their own words.

When a gap turns out to be a lead in disguise, route it into the flow described in using an AI chatbot for lead generation so the question ends in a captured enquiry rather than a dead end.

Keep the answers owned by someone

Assign one person to review FAQ answers when your services, hours, or policies change. Stale answers are the quiet failure mode of FAQ bots. The Christmas hours from two years ago will keep being quoted with total confidence until someone updates them.

Write answers from your real policies

Build each FAQ answer from the policy or page it describes, in the words your staff use on the phone. When the cancellation answer in chat matches what a customer hears when they call, nobody feels they were told one thing online and another in person. Where a policy has an exception that staff grant often, write the exception into the answer or route the question to a person. A bot that quotes a hard policy while staff bend it daily creates arguments at the front desk, with the chat transcript as evidence.

Seasonal and short lived answers

Holiday hours, temporary closures, and one off offers need answers with an end date, and a diary note to remove them. The safest pattern is a single list of temporary answers that one person checks at the start of each month. Temporary answers cause a special kind of damage because everything else the bot says is accurate, so customers believe the one stale answer completely. A bot that is right about parking, prices, and preparation, and wrong about being open on a holiday, has used its credibility to send a customer to a locked door.

Review the bot answers after any price or policy change

Tie FAQ updates to the changes that trigger them. When prices move, when a service is dropped, when the cancellation window changes, the chat answers change the same day, by the same person who made the change. A shared checklist on the wall or in the team chat is enough. The failure this prevents is subtle. Customers rarely complain that a bot quoted an old price. They simply arrive expecting the old price, and your staff absorb the awkward conversation that follows, one booking at a time.