What is a simple FAQ chatbot?
A chatbot that answers customer questions using only the document you gave it — not the open internet, not a guess. What it's good for, and why grounding it in your own content matters more than which AI model powers it.
The problem
The same handful of questions, asked over and over, on a website with no one watching it overnight
Most businesses have a short list of questions that come up constantly — shipping times, return policy, pricing, how something works. Answering them individually, every time, by email or live chat, is a poor use of a person's time, and outside business hours nobody is answering them at all, so the visitor either waits or leaves.
What it is
A chatbot that only knows what you told it
A simple FAQ chatbot is a web-embeddable chat widget backed by a document you provide — your actual FAQ page, policy document, or product information. When someone asks a question, the bot searches that document for the relevant section and answers from it, rather than answering from general internet knowledge.
This distinction matters more than it sounds. A chatbot answering from the open internet can confidently state something that isn't true about your specific business. A chatbot grounded in your document can only answer what's actually written there — and, built properly, will say it doesn't know rather than invent an answer.
A well-built one will:
- Answer strictly from the document you provided, not general knowledge
- Say plainly when a question isn't covered by that document
- Match your actual tone and terminology, not a generic customer-service voice
- Embed cleanly on your existing website without needing a developer to maintain it
The realistic goal: Handle the repetitive questions instantly, at any hour, so a human only gets pulled in for what genuinely needs a person.
Why it matters
Availability is part of the product now
A visitor with an unanswered question at 11pm is a visitor who often just leaves. Instant, accurate answers outside business hours capture interest that would otherwise evaporate by morning.
Repetitive questions are expensive precisely because they're repetitive. The hundredth person asking about your return policy deserves the same accurate answer as the first, without costing a person the same five minutes each time.
A grounded chatbot is safer than it sounds. Because it only draws from a document you control, updating the bot's knowledge is as simple as updating that document — no retraining, no waiting on a developer.
Best practices
Getting a chatbot that helps instead of frustrates
Write the source document for a chatbot, not just for humans
A well-organized FAQ with clear headers and direct answers gives the bot far better material to search than a single wall of marketing copy. Structure the document the way you'd want the answer phrased.
Explicitly instruct it to say 'I don't know' when it should
The single most important behavior to test for is what happens when a question genuinely isn't covered. It should say so and offer a way to reach a human, not stretch an unrelated answer to fit.
Test it with the weird questions first
Before launching, throw it edge cases — questions phrased oddly, questions slightly outside your FAQ's scope, questions in a different tone. That's where a poorly grounded bot reveals itself.
Keep the source document current
A chatbot is only as accurate as the document behind it. A stale FAQ produces a bot confidently repeating outdated shipping times or a discontinued policy.
Give it a visible way to hand off to a human
For anything sensitive — a complaint, a refund dispute, a question outside its scope — make the path to a real person obvious rather than trapping the visitor in a loop with the bot.
The mistake that costs the most: Feeding it a document that's out of date, then trusting it blindly because it "sounds right." A confident, well-phrased wrong answer is worse for trust than a visibly outdated FAQ page, because it doesn't look wrong.
Limits
What it will not do for you
It cannot answer questions about information that isn't in the source document — account-specific details, order status, anything that requires looking something up in a separate system.
It doesn't negotiate, discount, or make exceptions. Those are judgment calls for a person, not something to delegate to a document-grounded bot.
It won't replace a support team for genuinely complex issues — it's built for the repetitive, well-defined questions, not the edge cases that need real problem-solving.
Simple FAQ Chatbot — This guide covers what the chatbot does and how to prepare the document behind it. The Builder 1 session is the build — a Python and Streamlit app using the Claude API to answer questions grounded in an FAQ document you provide, deployable to any website.
Frequently asked questions
Can it answer questions that aren't in my FAQ document?
A well-built version deliberately won't — it's designed to answer only from the document you provide and say so plainly when something falls outside it, rather than guessing from general knowledge.
How do I update what the bot knows?
By updating the source document. There's no retraining involved — the bot searches the current version of the document each time it answers, so an edit takes effect immediately.
Can I embed it on any website, or only certain platforms?
The Builder 1 build produces a Streamlit app that can be embedded via an iframe or linked directly, which works on essentially any website regardless of what it's built on.
Is this the same as a general-purpose AI chatbot like ChatGPT?
No — the key difference is grounding. A general AI chatbot draws on broad internet knowledge; this one is restricted to your document, which makes its answers specific and accurate to your business rather than generic.
What happens if someone asks something inappropriate or off-topic?
A well-prompted bot stays within its lane and declines to engage with off-topic requests, redirecting the visitor back to what it can actually help with.