The questions customers ask a chatbot fall into eight families: pricing and quotes, delivery and lead times, opening hours and availability, "will this work for my situation", comparison with a competitor, problems after purchase, admin (invoices, cancellation, GDPR) and requests to talk to a person. For each family, a chatbot only answers correctly if the answer exists in writing in its knowledge base: pricing page, terms, product pages, procedures.
This guide takes the visitor's point of view. It lists the questions you can expect, sector by sector, and says which document to add to the knowledge base for each. It also covers the questions where the chatbot should hand over, and a method to catch the ones that stay unanswered after launch.
Note: every question quoted below is a typical example written for this article. None is an excerpt from a real conversation, and no figure here comes from real usage data.
If you are thinking about retiring a static FAQ, read why nobody reads your FAQ page. Here the angle is different: we start from what visitors ask, not from what you want to say.
Table of Contents
The 8 families of questions and the content to prepare
Visitors do not phrase questions like your navigation menu. They write "how much for three people?" or "do you ship to Ireland?", not "Pricing". A chatbot connected to your content (RAG) retrieves the relevant passages, provided they exist and are clear.
The table summarises the eight families, with example questions and the content to load into the knowledge base for each.
| Family | Example questions | Content to put in the knowledge base |
|---|---|---|
| Pricing and quotes | "How much does it cost?" "Do you give free quotes?" "Are there hidden fees?" | Up-to-date pricing page, price ranges, how a quote works, what is included or excluded |
| Delivery, lead times, coverage | "Do you ship to Ireland?" "How long for an order placed today?" "Do you cover my area?" | Shipping policy, lead time and cost table by region, list of towns or postcodes covered |
| Availability and hours | "Are you open on Saturday?" "Is this model in stock?" "When can I come in?" | Opening hours and closing days, availability rules, booking procedure |
| "Will it work for my situation?" | "Is it compatible with my setup?" "Is it suitable for a small business?" | Detailed product or service pages, compatibility criteria, typical use cases, stated limits |
| Comparison with a competitor | "How are you different from X?" "Why choose you?" | Factual positioning page, verifiable differences, cases where your offer is not the right fit |
| Problem after purchase | "My parcel arrived damaged" "How do I reset the device?" "Warranty?" | Manuals, troubleshooting steps, warranty and returns terms, complaints process |
| Admin | "Where is my invoice?" "How do I cancel?" "What do you do with my data?" | Terms and conditions, privacy policy, cancellation procedure, legal notices |
| Talk to someone | "I want a human" "Can someone call me back?" | Contact details, contact hours, rule for handing over to a form, a ticket or a booking |
Pricing and quotes. This is often the first question on a site. If your pricing page is vague, the chatbot will be too. Add ranges, the factors that move the price and the quote process. A chatbot that says "from" with the right conditions beats silence.
Delivery and coverage. Answers must be factual and dated: lead times, working days, countries served. "Fast delivery" answers nothing. A lead-time table by region does.
Availability. Hours change (bank holidays, summer closures). Keep one document you update, rather than hours repeated across ten pages that the chatbot could mix up.
Fit and edge cases. These are the commercially richest questions. They need precise pages and examples. Also write down what you do not do: that stops the chatbot from inventing a promise.
Comparison. Do not let the chatbot improvise about competitors. Write a factual page on your real differences and tell the agent in its instructions to stick to your own offer.
After-sales. Precision matters most here. Manuals, numbered troubleshooting steps and warranty terms belong in the base. Our guide to ticket deflection covers the full support flow.
Admin. Terms, privacy policy and cancellation procedure are documents many sites never reread. The chatbot will quote them to the letter, so review them before importing.
Typical questions by sector
The families are the same everywhere, but the wording changes. Here are typical examples for five sectors, to adapt to your business.
E-commerce
- "Where is my order?"
- "Can I return an item I've worn once?"
- "How long does delivery to Germany take?"
- "Is this in a size 10?"
- "Can I pay in instalments?"
- "The newsletter discount code isn't working."
- "How do I get a VAT invoice?"
Prepare: terms of sale, returns policy, shipping table, product pages, payment FAQ. For order tracking, set a clear instruction (tracking link, form) instead of letting the chatbot guess.
SaaS
- "What is the difference between the Starter and Pro plans?"
- "Can I try it without a credit card?"
- "How do I connect your tool to my CRM?"
- "Where is my data hosted?"
- "How do I cancel my subscription?"
- "Is there an API?"
- "My account is locked. What now?"
Prepare: pricing page, product documentation, integrations pages, privacy policy, cancellation procedure, onboarding guide.
Contractors and home services
- "Do you cover my postcode?"
- "How much is an emergency call-out on a Sunday?"
- "Do you give free quotes?"
- "When can you come this week?"
- "Are you insured?"
- "Can I send photos of the problem?"
Prepare: coverage list, price ranges, emergency terms, insurance details, a rule for handing over to a contact form or a booking. See also our guide on qualification questions before a site visit.
Professional practices (clinics, law firms, accountants)
- "How do I book an appointment?"
- "What should I bring to the first consultation?"
- "What are your fees for an initial meeting?"
- "Do you accept my insurance?"
- "Can I cancel or reschedule?"
- "Do you see clients in person or by video?"
Prepare: fee schedule, booking terms, list of documents to bring, cancellation policy. The chatbot explains how the practice runs. It gives no legal advice and no medical advice (see the next section).
Hotels and restaurants
- "What time is breakfast?"
- "Is there parking, and is it paid?"
- "Are dogs allowed?"
- "Do you have a table for six on Saturday night?"
- "Can I cancel the booking for free?"
- "Do you serve gluten-free dishes?"
- "How do I get to you from the station?"
Prepare: a practical sheet for the venue (hours, services, parking, pets), cancellation terms, an up-to-date menu with allergens, directions. For more, see our hotel chatbot guide.
Questions the chatbot should not answer alone
A good chatbot knows how to say "I'll pass this on". Some questions should not get an automated answer, because they engage your responsibility or need human judgement. The rule: the chatbot informs, a person decides.
| Type of question | Why hand over | Relay to plan |
|---|---|---|
| Personalised legal, medical or financial advice | The chatbot does not know the full situation and cannot commit the professional | Contact form or appointment booking |
| Custom quote, exceptional discount | Depends on your commercial decision | Contact form that collects the need and contact details |
| Dispute, complaint, refund | Requires checking the customer's real file | Support ticket |
| A specific customer's personal data (an exact order, a contract) | Needs reliable identification | Ticket, or identified visitor if your site supports it |
| Emergency (breakdown, safety, health) | Human response time matters | Emergency number displayed, form flagged as urgent |
| Explicit request for a human | The visitor asked for one | Form, ticket or booking depending on the case |
In Heeya, these relays are agent tools: a contact form to capture a request (Standard and Premium plans), support tickets and appointment booking with slots and email confirmation (Premium plan). You decide in the agent's instructions when to offer them. Our chatbot solution page presents the whole set. For when and how to hand over, read our chatbot human handoff guide.
Method: collect, test, monitor
1. Start from your own sources of questions
You do not need to guess the questions: they already exist in your business. The best sources:
- Your contact inbox: reread the last 50 messages and group them by family.
- Phone calls: ask whoever picks up to note recurring questions for two weeks.
- Existing support: tickets, customer reviews, social media messages.
- Your site's internal search: the terms typed in the search bar (if your tool logs them) show what visitors cannot find.
- Google Search Console: the queries that bring visitors show their natural phrasing.
Once the list exists, write each answer in a clear document, or import your existing question-and-answer pairs. Heeya accepts Q&A imports in xlsx or csv, in addition to PDF, DOCX, PPTX, TXT and web page scraping. For content structure, read our guide on knowledge base engineering for AI chatbots, and try the RAG knowledge base readiness assessment.
2. Test before going live
Ask your own 30 to 50 typical questions, rephrased like a customer in a hurry: typos, short sentences, non-technical words. For each answer, note whether it is right, incomplete or wrong. Have the person who usually answers customers review the results.
Also test accidental traps: an off-topic question, a request for a discount, a question about a product you do not sell. The chatbot should decline politely or hand over, not invent.
3. Monitor unanswered questions after launch
Content is never complete on launch day. In Heeya, the Analytics tab shows each conversation with a reliability indicator for every answer. When the retrieved content looks insufficient, the answer is flagged with the conditional wording "information probably not found". It is an indicator, not proof: it tells you where to look first.
The useful routine: each week at first, scan the flagged answers, identify the question, then complete the base (a new paragraph, a new page, or an imported Q&A pair). After a few weeks, the pace slows down.
Chatbot knowledge base checklist
Tick as you go (ticks are not saved; this is a reminder for the current page).
Prepare your knowledge base with Heeya
Import your documents, test your typical questions and track answer reliability. 7-day free trial, no card required.
FAQ: questions customers ask a chatbot
What questions do customers ask a website chatbot?
They fall into eight families: pricing and quotes, delivery and lead times, opening hours, fit for their situation, comparison with a competitor, after-sales problems, admin (invoice, cancellation, GDPR) and requests for a human. The exact wording depends on your sector.
How many questions should I prepare before launch?
There is no magic number. About thirty typical questions covering all families are enough for a first test. Document quality matters more: one clear pricing page beats ten poorly documented questions.
Where can I find my customers' real questions?
In your contact inbox, phone calls, support tickets, reviews, your site's internal search and Google Search Console queries. These sources show the exact words your visitors use.
What should go in a chatbot knowledge base?
The documents that already contain the answers: pricing page, terms, product or service pages, manuals, after-sales procedures, hours, privacy policy. You can also import question-and-answer pairs in xlsx or csv. See our FAQ prompt guide for how to write them.
Can a chatbot answer pricing questions and give quotes?
It can give your public prices and ranges and explain how a quote works. For a custom quote, a contact form that collects the need and passes it to a person is safer.
Which questions should a chatbot not answer alone?
Personalised legal, medical or financial advice, disputes and refunds, a specific customer's personal data and emergencies. Plan a relay for these: a form, a support ticket or an appointment.
How do I find which questions are left unanswered?
By reviewing conversations regularly. In Heeya, the Analytics tab gives a reliability level for each answer and flags those where the information was probably not found in the base. You then complete your documents.
Should I replace my FAQ page with a chatbot?
Not necessarily. A good FAQ remains useful for search and as a source for the chatbot. Our article on replacing your FAQ page explains why a chatbot is a useful complement to an FAQ few visitors read.
Can the chatbot answer in several languages?
The Heeya widget follows the page language, in French and English. If your site is bilingual, prepare the key documents in both languages.
Conclusion
Preparing a chatbot's answers starts with listening to what visitors already ask. The eight families cover most of it: if each one has a clear document in the base and a hand-over rule, the chatbot answers correctly on the bulk of your traffic.
The rest is settled over time: test before launch, monitor after, complete weekly at first. A chatbot is a knowledge base that improves with real questions, not a project that ends on launch day.
