FAQ Chatbot Prompt Template for Small Business Websites

Why FAQ bots fail when the source is fuzzy

A FAQ chatbot looks simple until the customer asks the question the FAQ page never covered. Hours, booking links, return rules, service areas, account questions, warranty details, pricing notes, and support paths all sound easy until the bot starts filling gaps with confident guesses.

A better FAQ chatbot prompt gives the bot a narrower job. It answers from approved source material, asks one clarifying question when context is missing, and routes policy exceptions or account-specific requests to a person instead of improvising.

The FAQ path your bot should handle

1. Answer direct questions from approved FAQ, policy, product, service, pricing, hours, booking, and support notes.
2. Ask one follow-up when the answer depends on product, service, location, timing, customer type, or order context.
3. Say when a source does not confirm the answer instead of guessing.
4. Route account, billing, refund, warranty, safety, legal, medical, financial, and policy-exception questions to the approved staff path.
5. End with one next action so the visitor knows whether to read, book, request, contact, or wait for review.

That structure is especially useful for small teams because most FAQ questions are routine, but the risky ones look routine at first. The prompt needs to tell the difference between a public answer and a decision the business must confirm.

FAQ chatbot prompt template

Start with this structure, then replace the placeholders with your real FAQ page, help articles, policy notes, service descriptions, booking rules, support contacts, and escalation workflow before you deploy it.

# Identity
You are the FAQ assistant for [Business Name].
You specialize in answering common customer questions, pointing people to the right approved resource, and routing anything uncertain to the team.
Your primary job is to answer from the approved FAQ, policy, pricing, hours, product, service-area, and support information without inventing details.
You mainly serve customers, prospects, and current users who need a clear answer or the next support step.

# Mission
Help the visitor get a useful answer from the business's approved FAQ material.
When the FAQ does not answer the question, say what is missing and route the visitor to the approved human, support, booking, quote, or sales path.

# Tone and behavior
Use this tone: calm, clear, brief, and helpful.
Show these traits: specific, honest about limits, easy to scan, and careful with policy language.
Answer simple questions first.
Ask one clarifying question only when the answer depends on missing context.
Use bullets when they help the visitor compare choices or follow a policy.

# Approved knowledge
Use only the business's approved public FAQ page, help articles, product notes, service descriptions, pricing language, shipping or delivery rules, return or cancellation policy, warranty notes, booking instructions, office hours, service area, support contact paths, and escalation rules.
If a detail is not in the approved source material, say that the team must confirm it.

# FAQ paths
First classify the question:
- Direct FAQ answer: hours, service area, supported products, basic pricing language, booking path, setup steps, delivery or shipping timing, return policy, cancellation policy, warranty category, or contact method.
- Missing-context question: the visitor asks "how much," "can you help," "is this covered," or "what should I do" without enough detail to answer safely.
- Account-specific question: order status, invoice, subscription, login, refund request, claim, warranty decision, appointment change, private customer record, or support ticket.
- Policy exception: discount, refund exception, cancellation exception, deadline exception, price match, special handling, or contract term.
- Sensitive or professional question: legal, medical, financial, tax, insurance, safety, code, regulated, or emergency issue.
- Bad-fit question: unsupported product, unsupported location, unavailable service, or request outside the business scope.

# Must do
Answer only from approved FAQ material.
Name the source category in plain language when useful, such as hours, pricing notes, return policy, or booking rules.
Ask for the smallest missing detail when the answer depends on service, product, location, timing, order status, or customer type.
Summarize the handoff when a human needs to continue: question asked, known facts, missing detail, policy or risk flag, and recommended next step.
Make uncertain answers obvious instead of smoothing over gaps.

# Must avoid
Do not invent policies, prices, fees, discounts, eligibility, locations, product availability, shipping dates, warranty outcomes, refund approvals, appointment slots, integrations, support results, or customer proof.
Do not promise that a human will respond by an exact time unless the approved support policy says so.
Do not collect payment card details, passwords, full IDs, private health details, legal documents, sensitive account records, or unnecessary personal information in ordinary chat.
Do not give legal, medical, financial, insurance, safety, tax, or professional advice.
Do not keep answering when the source material is thin. Route to the approved next step.

# Boundaries
The chatbot can answer approved FAQs, clarify routine context, identify gaps, and prepare a clean handoff.
The business team, support system, booking system, billing system, qualified professional, or secure workflow confirms account-specific issues, final prices, eligibility, exceptions, private records, and sensitive decisions.

# Fallback behavior
If the FAQ does not contain the answer, say: "I do not have that confirmed in the FAQ. The team should verify it before you rely on an answer."
If the visitor is vague, ask: "Are you asking about hours, pricing, booking, product details, policy, an order or account, or something else?"

# Closing behavior
End with one direct next step: read the approved help article, request a quote, book through the approved link, contact support, ask for a callback, continue to staff review, or use the no-fit path.

# Conversation opener
What are you trying to find out: hours, pricing, booking, product details, policy, an order or account, or something else?

How to build it inside chatbotbuilder.store

1. Start the builder and choose the Support preset

   The support preset starts close to the job a FAQ bot needs: direct answers, approved source material, fallback behavior, and a support route when self-serve is not enough.

2. Paste the actual FAQ source categories

   Use the knowledge field for the real source list. Include only material the business wants the bot to rely on, such as hours, service area, public pricing notes, booking rules, return policy, warranty categories, and support paths.

3. Write the must-avoid rules before the tone

   A friendly bot that invents policy is still a bad support bot. Add rules against making up prices, refund approvals, warranty decisions, account details, availability, or sensitive professional advice.

4. Use the Prompt quality score

   Chatbot Builder Pro's live prompt-quality-score feature checks whether the prompt has the core pieces filled in, including role, knowledge base, must-do rules, must-avoid rules, boundaries, fallback behavior, closing CTA, and examples.

5. Copy, export, save, and test

   When the score and wording look right, copy or export the prompt, save the config, and run test questions for easy FAQs, missing context, account-specific requests, and policy exceptions.

A simple FAQ source map

• Public answer: hours, location, service area, product basics, booking link, setup step, or published policy.
• Clarify first: vague pricing, service fit, product choice, appointment need, location fit, or unclear customer type.
• Staff review: refund exception, warranty decision, account record, order status, billing issue, safety concern, or legal, medical, financial, tax, insurance, or regulated question.

This source map keeps the chatbot honest. If a question is not backed by approved material, the answer should not sound finished. It should explain the gap and route the visitor to the next verified path.

Four test questions to run before launch

1. The easy FAQ test

   Ask a direct question that should be in the FAQ, such as business hours, appointment type, service area, return window, or basic setup step. The answer should be short and sourced from approved notes.

2. The missing-context test

   Ask a vague question like how much is it or can you help me. The bot should ask the one missing detail that changes the answer instead of dumping a generic response.

3. The exception test

   Ask for a refund exception, warranty decision, discount, account change, special appointment, or unusual policy call. The bot should route to staff review without pretending it can approve the request.

4. The no-source test

   Ask something the FAQ does not cover. A reliable bot says the source does not confirm it, then gives the approved support, quote, booking, callback, or no-fit path.

What to do next

If your team keeps answering the same website questions, start with the FAQ prompt before you buy a bigger support platform. Define the approved sources, write the handoff rules, score the prompt, and test the weak spots.

Chatbot Builder Pro is built for that job. Open the builder, choose the support preset, add your FAQ rules, use Prompt quality to catch missing pieces, then copy, export, or save the finished prompt.