Table of Contents
- Key takeaways
- What is an FAQ chatbot?
- FAQ chatbot examples: what the two types do with the same questions
- How to choose a way to build an FAQ chatbot
- Option 1: Script the answers in a flow builder
- Steps
- Cost
- Limits
- Who it fits
- Option 2: Train a no-code chatbot platform on your existing content
- Steps
- Cost
- Limits
- Who it fits
- Option 3: Turn on the AI agent in your helpdesk
- Steps
- Cost
- Limits
- Who it fits
- Option 4: Build an FAQ chatbot on the OpenAI API
- Steps
- Cost
- Limits
- Who it fits
- Should the chatbot go on the FAQ page?
- How to keep an FAQ chatbot accurate
- Common mistakes when building an FAQ chatbot
- Frequently asked questions
- Basics
- What is an FAQ chatbot?
- How do you build an FAQ chatbot?
- What is the difference between an FAQ chatbot and an FAQ page?
- Can an FAQ chatbot answer questions that are not on the FAQ page?
- Building and training
- What content should an FAQ chatbot be trained on?
- How long does it take to build an FAQ chatbot?
- Do you need to write question and answer pairs?
- Can ChatGPT be used as an FAQ chatbot?
- Cost
- Is there a free FAQ chatbot?
- How much does an FAQ chatbot cost?
- Which option has the most predictable cost?
- Accuracy and upkeep
- How do you stop an FAQ chatbot from giving wrong answers?
- How do you update an FAQ chatbot when the answers change?
- Can an FAQ chatbot hand a conversation to a person?
- Sources
Created by
Do not index
Created time
May 28, 2026 11:20 AM
An FAQ chatbot is a chat widget that answers common customer questions on a website, and you can build one in four ways: script the answers in a flow builder, train a no-code platform on your existing content, turn on the AI agent in your helpdesk, or code it on the OpenAI API.
TL;DR
- A scripted FAQ bot matches a question to an answer you wrote. A content-trained FAQ chatbot writes each answer from your help pages, so it also covers questions nobody wrote down.
- Under about 20 fixed questions, a scripted flow is enough. Tidio's Free plan costs $0 and reaches 100 visitors a month with Flows.
- For answers drawn from a help center, website, or PDFs without code, use a content-trained platform. SiteGPT Starter costs $59 a month billed monthly, or $39 a month billed yearly, after a 7-day trial that needs a card.
- Teams that already run a helpdesk can add its AI agent. Fin costs $0.99 per outcome with a minimum of 50 outcomes a month.
- Developers can build one on the OpenAI API. File search costs $2.50 per 1,000 calls plus model tokens. That comes to about $2.95 for 1,000 questions on the cheapest listed model.
Key takeaways
Option | Best for | Cost to start | Main limit |
1. Scripted FAQ flow (Tidio Flows) | A short, fixed list of questions | $0 on Tidio's Free plan | Answers only the questions you scripted |
2. Content-trained platform (SiteGPT) | Answers from existing docs, with no code | $59 a month billed monthly, or $39 a month billed yearly | No free plan, and monthly message quotas |
3. Helpdesk AI agent (Fin) | Teams that already work in a helpdesk | $0.99 per outcome, 50 outcomes a month minimum | Cost rises with every resolved conversation |
4. OpenAI API build | Developers who want the answers inside their own product | Usage-based: $2.50 per 1,000 file search calls plus tokens | Needs code, hosting, and upkeep |
What is an FAQ chatbot?
An FAQ chatbot is a chatbot that answers the questions customers ask most often, such as prices, shipping times, refunds, and account setup. It sits on a website or inside a messaging channel and replies at any hour.
Two different products share the name. The difference decides how much work the build takes and how many questions the bot can answer.
Type | How it answers | What you maintain | What happens with a new question |
Scripted (rule-based) FAQ bot | Matches a button click or keyword to an answer you wrote | Every question and answer pair | No answer until you add one |
Content-trained (AI) FAQ chatbot | Finds the relevant passages in your content and writes an answer from them | The source content | Answers if the content covers it, and says it does not know if not |
A content-trained chatbot uses retrieval. The model's weights do not change. The system searches an index of your content for each question. The longer explanation is in what is RAG.
FAQ chatbot examples: what the two types do with the same questions
FAQ chatbot examples are easiest to judge by putting the same visitor questions to both types. The store below is hypothetical. It has a 12-question FAQ page, a shipping policy page, and a pricing page.
Visitor question | Scripted FAQ bot | Content-trained FAQ chatbot |
"What is your return policy?" | Answers. The question is on the list. | Answers from the returns page. |
"Can I send back a sale item I bought three weeks ago?" | Shows the general returns answer, or nothing. | Answers if the returns page covers sale items and time limits. |
"Do you ship to Norway, and what does it cost?" | Answers only if someone scripted Norway. | Answers from the shipping policy page. |
"Is the second seat free on the middle plan?" | No match. | Answers from the pricing page if the seat rule is written there. |
"Where is my order?" | Shows the tracking link you scripted. | Explains how to track an order. It cannot look up the order without a connection to the order system. |
"Can I speak to someone?" | Hands off if the flow has that step. | Hands off if human escalation is turned on. |
The pattern holds across most sites. A scripted bot is reliable for the questions on the list. A trained chatbot covers the long tail of questions that the content answers but the list does not.
To see FAQ bots that organizations run in public, including USCIS Emma and the IRS chatbots, read the FAQ and customer service chatbot examples.
How to choose a way to build an FAQ chatbot
Choose by three things: how many distinct questions customers ask, where the answers already live, and who will maintain the bot.
If you need... | Pick... | Why |
Answers to fewer than 20 fixed questions, at no cost | Option 1: a scripted flow | Every answer is written and approved by you |
Answers from a help center, website, or PDFs that already exist | Option 2: a content-trained platform | No question and answer pairs to write, and no code |
Exact wording for a refund or legal answer | Option 1, or a custom response in option 2 | A script returns the same text every time |
The bot inside the helpdesk your agents already use | Option 3: the helpdesk AI agent | Handoffs land in the same inbox |
Answers inside your own app or product | Option 4: an OpenAI API build | Full control over the interface and the data flow |
A fixed monthly bill | Option 2 | Plans are flat. Options 3 and 4 bill by usage |
A free FAQ chatbot | Option 1 | Free plans exist for scripted flows. The other options are paid |
The four options below follow one template: steps, cost, limits, and who it fits. No option is best for every team.
Option 1: Script the answers in a flow builder
Best for: a short, fixed list of questions.
Scripting the answers in a flow builder means you write each question and answer yourself and connect them in a visual editor. The visitor clicks a button or types a keyword, and the bot returns the matching answer.
Tidio is the example here. Its pricing page lists a visual Flows builder and FAQ Flows templates. If you want to compare canvas tools before you pick one, the drag and drop chatbot builder guide ranks seven of them.
Steps
- List the 10 to 20 questions customers ask most. Pull them from the customer service inbox.
- Write one short answer for each question and get it approved.
- Create a Tidio account and open the Flows builder.
- Start from an FAQ template, then replace the sample questions with your own.
- Add a final branch that passes the visitor to a person or collects an email address.
- Install the widget and test every branch.
Cost
Tidio item | Price on the pricing page | What it includes |
Free plan | $0 | 50 billable conversations and 100 visitors reached by Flows a month |
Flows | From $24.17 a month, billed annually | From 2,000 visitors reached a month |
Lyro AI Agent | From $32.50 a month, billed annually | From 50 Lyro AI conversations a month. The first 50 are free, once |
At the annual rate, the Flows plan costs $290.04 a year.
Limits
- The bot answers only the questions you scripted. Tidio's own pricing FAQ says Flows follow set rules and pathways.
- Each new question, price change, or policy change means another manual edit.
- A long list turns into a menu that visitors have to scan, the same as a static FAQ page.
- The Free plan reaches 100 visitors a month with Flows.
Who it fits
A scripted flow fits a small site with a short list of stable questions and no budget. Other builders in this category are compared in best no-code chatbot builders, and free plans are covered in free chatbots for a website.
Option 2: Train a no-code chatbot platform on your existing content
Best for: answers from existing docs, with no code.
Training a no-code chatbot platform means you point it at the content you already have, and the platform indexes it and answers from it. You do not write question and answer pairs.
SiteGPT is the worked example here. The steps use the menu names from its documentation.
Steps
- Start the trial and create a chatbot. The 7-day trial needs a credit card. Sign in with the email link, select Create New Chatbot, and name it.
- Add the pages that already answer questions. Select Website Links, then Scrape Website or Add from Sitemap. Start with the FAQ page, the help center, the pricing page, and the shipping or policy pages.
- Add files and connected apps. Under Files & Data Sources, upload files such as PDFs, or connect Notion, Google Drive, Dropbox, OneDrive, or Box. Keep each file under 10 MB. Scanned PDFs give no text, because SiteGPT does not run OCR.
- Write the instructions. Select Chatbot Instructions, then Add New Instructions. Write one rule per line. A starting set is below.
- Pick a persona. There are 12 built-in personas. Default keeps answers to 1 to 3 sentences. That length suits FAQ answers. If you want to define a custom tone and write the welcome message before you pick, the chatbot design guide covers persona and opening-turn patterns.
- Add custom responses for answers that must be right. Under Custom Responses, write the question and the answer for prices, refunds, and policies. SiteGPT matches them by meaning, so "How much does it cost?" also covers "What's the price?".
- Test before launch. Select Chat with your chatbot on the Dashboard. Ask the last 20 real customer questions, plus a few the content does not cover. The chatbot should reply with the fallback message for those.
- Install it. Select Installation, copy the JavaScript Embed code, and paste it before the closing body tag. WordPress sites can use the SiteGPT plugin with the chatbot ID.
- Turn on human support. Under Leads, select Human Support and turn on Enable Human Support. It is available on every plan.
- Review Chat History every week. Filter by negative reactions, open the wrong answer, and correct it with Incorrect? Modify bot's answer!
Instructions to start from:
You are the customer service assistant for [Company Name].
Answer only from the content you were given.
Give the direct answer first, in one to three sentences.
If the content does not answer the question, say so and offer a person.
Link the help page the answer came from when there is one.If the bot does more than FAQs, such as product information or order tracking, copy a longer instruction set from the chatbot template page. How to prepare the content itself is covered in how to train a chatbot on your own data. For product documentation sites, see AI chatbots for documentation.
Cost
SiteGPT plan | Billed monthly | Billed yearly | Messages a month | Pages of content | Content refresh |
Starter | $59 a month | $39 a month ($468 a year) | 4,000 GPT-4.1-mini or 400 GPT-4.1 | 1,000 | Manual |
Growth | $129 a month | $79 a month ($948 a year) | 10,000 GPT-4.1-mini or 1,000 GPT-4.1 | 10,000 | Monthly |
Scale | $429 a month | $259 a month ($3,108 a year) | 40,000 GPT-4.1-mini or 4,000 GPT-4.1 | 50,000 | Weekly, with a daily sitemap scan |
Yearly billing on Starter costs $468 against $708 for twelve monthly payments. That saves $240 a year, or 34%.
Limits
- There is no free plan. The 7-day trial needs a credit card.
- Message quotas are fixed per plan. A GPT-4.1 message uses 10 times the quota of a GPT-4.1-mini message.
- Starter refreshes content by hand and has no chat integrations or API access.
- Scheduled refresh covers website crawls, sitemaps, and GitHub. Uploaded files and connected apps such as Notion and Google Drive need a manual resync. If the FAQs live in Notion and change every week, the Notion chatbot guide compares tools that re-read Notion on a schedule.
- A custom response is preferred, but the chatbot can reword it. An instruction to use it word for word makes exact wording more likely. It is not a guarantee.
- The Unanswered questions list in Analytics needs Growth or above.
- The rule to stay inside your content is an instruction to the model. SiteGPT's documentation calls it strong but not absolute.
Who it fits
A content-trained platform fits a business that already has help content and has nobody to build or maintain a bot. It fits less well when every answer must be word for word, or when the budget is zero. If the chatbot must also carry your brand and run on channels beyond the website, how to build a custom AI chatbot covers those layers.
Option 3: Turn on the AI agent in your helpdesk
Best for: teams that already work in a helpdesk.
Turning on the AI agent in your helpdesk means the FAQ chatbot runs inside the tool your agents already use. It reads the help center, answers in the messenger, and passes the conversation to an agent in the same inbox.
Fin is the example here. Fin runs on the Intercom helpdesk, and its pricing page says it also works with other helpdesks, including Salesforce and HubSpot.
Steps
- Check that the help center is current. The agent answers from it.
- Turn on the AI agent and connect the help center and other knowledge sources.
- Set the tone and the answer length.
- Set the handoff rules, so a person takes over when the agent cannot resolve the question.
- Test with real past tickets before it goes live for all visitors.
- Set a usage alert or limit, because billing follows outcomes.
Cost
Fin setup | Price on the pricing page | Monthly floor |
Fin with your current helpdesk | $0.99 per outcome, 50 outcomes a month minimum | $49.50 |
Fin with the Intercom helpdesk | From $0.99 per outcome, plus a helpdesk seat from $19 a month | Depends on seats and outcomes |
The cost scales with volume. At $0.99 per outcome, 500 outcomes cost $495 a month and 1,000 outcomes cost $990 a month.
Limits
- The bill grows with every conversation the agent resolves. A busy month costs more than a quiet one.
- Fin counts an outcome when it resolves the issue end to end, or when it completes a procedure that ends in a handoff.
- It adds the most value when the team already uses a helpdesk. Without one, the helpdesk is a second purchase.
Who it fits
A helpdesk AI agent fits a customer service team that already lives in a helpdesk and wants one inbox for bot and human conversations. Teams on Zendesk can compare the two approaches in SiteGPT vs Zendesk.
Option 4: Build an FAQ chatbot on the OpenAI API
Best for: developers who want the answers inside their own product.
Building an FAQ chatbot on the OpenAI API means writing code that stores your files in a vector store and lets a model search them before it answers. OpenAI hosts this as the file search tool in the Responses API.
Steps
- Create an OpenAI API account and an API key.
- Export the FAQ page and help articles as files.
- Create a vector store and upload the files to it.
- Call the Responses API with the
file_searchtool and the vector store ID.
- Add a system instruction to answer only from the files and to say so when they have no answer.
- Build the chat interface, host it, and add a way to re-upload files when the content changes.
Cost
The API bills by usage. File search costs $2.50 per 1,000 tool calls. Storage costs $0.10 per GB per day, and the first GB is free. Model tokens are billed on top.
Worked example. Assume 1,000 answered questions a month. Assume each question makes one file search call and uses 3,000 input tokens and 300 output tokens. Storage stays under the free 1 GB.
Cost line | On gpt-6-luna ($0.10 input, $0.50 output per 1M tokens) | On gpt-6.1-sol ($2.00 input, $10.00 output per 1M tokens) |
File search: 1,000 calls | $2.50 | $2.50 |
Input: 3M tokens | $0.30 | $6.00 |
Output: 0.3M tokens | $0.15 | $3.00 |
Total a month | $2.95 | $11.50 |
The token counts are assumptions. The totals leave out developer time, hosting, and the chat interface. To budget those developer hours and compare them with a platform plan over one and three years, see the cost to build a chatbot.
Limits
- It needs code. There is no widget, dashboard, or handoff to a person unless you build them.
- Content refresh is your job. Nothing re-crawls a website for you.
- Someone has to own it after launch.
Who it fits
An API build fits a team with developers that wants FAQ answers inside its own app. The ChatGPT and OpenAI routes are compared in more detail in how to train ChatGPT on your own data.
Should the chatbot go on the FAQ page?
Yes, put the chatbot on the FAQ page, and keep the page. The two do different jobs.
- Keep the FAQ page. It is training content for a content-trained chatbot, and search engines can read it.
- Add the widget to it. Visitors on an FAQ page are already looking for an answer. A chat box lets them ask in their own words.
- Add it to the pricing and checkout pages too. Questions about cost and delivery come up there.
- Use chat transcripts to update the page. A question that appears every week belongs on the page.
An embed script adds the chat icon to every page that carries the script. To limit the chatbot to the FAQ page, add the script to that page only. If you want ChatGPT itself on the page, read how to embed ChatGPT in a website to see which routes work.
How to keep an FAQ chatbot accurate
Keeping an FAQ chatbot accurate is mostly content work. Four habits cover it.
- Fix the source first. A wrong answer usually means the content is missing, outdated, or contradicted by another page. Correct the page, then refresh it.
- Keep an "I do not know" rule. Without it, a model fills gaps from general knowledge. The other guardrails are in the guide to chatbot hallucinations.
- Refresh on a schedule. A bot repeats whatever its sources say, including last year's prices.
- Read transcripts weekly. Unanswered questions show where the content has gaps.
The wider checklist, including disclosure and handoff, is in chatbot best practices.
Common mistakes when building an FAQ chatbot
The most common mistakes come from treating the chatbot as a one-time setup. Five show up often.
- Scripting 100 questions. Past about 20, a flow becomes hard to maintain and hard for visitors to use.
- Training on a stale help center. The chatbot repeats outdated answers with confidence.
- Hiding the route to a person. Visitors who cannot reach anyone leave or complain.
- Skipping the test set. Ask real past questions before launch, including ones the content does not cover. For a scored example with separate pass rates for answers, refusals, and handoffs, see the guide to chatbot training.
- Expecting account lookups from an FAQ bot. Order status and billing changes need a connection to the system that holds the data. That is the step from an FAQ chatbot to a full customer service chatbot.
Frequently asked questions
Basics
What is an FAQ chatbot?
An FAQ chatbot is a chatbot that answers the questions customers ask most often. A scripted one returns answers you wrote. An AI one writes answers from your help content.
How do you build an FAQ chatbot?
You build an FAQ chatbot in one of four ways: script the answers in a flow builder, train a no-code platform on existing content, turn on a helpdesk AI agent, or code it on the OpenAI API. The first two need no code.
What is the difference between an FAQ chatbot and an FAQ page?
An FAQ page is a list the visitor scans. An FAQ chatbot takes the question in the visitor's own words and returns one answer. Most sites keep both.
Can an FAQ chatbot answer questions that are not on the FAQ page?
A content-trained chatbot can, if another page or file it was trained on covers the question. A scripted bot cannot.
Building and training
What content should an FAQ chatbot be trained on?
Train it on the FAQ page, the help center, the pricing page, and the shipping, returns, and policy pages. Add internal documents only if they contain nothing private, because anyone who chats with the bot can reach that content.
How long does it take to build an FAQ chatbot?
It depends on the option and the content. A no-code platform has to crawl and index the pages before testing can start. Plan most of the time for testing with real questions.
Do you need to write question and answer pairs?
Only for a scripted bot. A content-trained chatbot answers from pages and files. A few written answers still help for prices and policies that must be right.
Can ChatGPT be used as an FAQ chatbot?
ChatGPT runs inside the ChatGPT app, so it does not sit on your website. A website FAQ chatbot uses an OpenAI model through a platform or the API. The routes are compared in how to train ChatGPT on your own data. If you are working from an older GPT-3 tutorial, read what a GPT-3 chatbot means today first, because OpenAI has shut those models down.
Cost
Is there a free FAQ chatbot?
Yes, for scripted flows. Tidio's Free plan costs $0 and includes 100 visitors reached by Flows a month. SiteGPT has no free plan. It has a 7-day trial that needs a card.
How much does an FAQ chatbot cost?
A scripted flow starts at $0. SiteGPT starts at $59 a month billed monthly, or $39 a month billed yearly. Fin costs $0.99 per outcome with a 50-outcome monthly minimum, so the floor is $49.50 a month. An OpenAI API build costs about $2.95 in usage for 1,000 questions on the cheapest listed model, plus developer time.
Which option has the most predictable cost?
A flat plan does. A platform plan stays the same each month up to its quota. Per-outcome and per-token pricing rise with volume.
Accuracy and upkeep
How do you stop an FAQ chatbot from giving wrong answers?
Tell it to answer only from your content and to say when it does not know. Keep the content current, and write fixed answers for prices and policies. No setting removes wrong answers completely.
How do you update an FAQ chatbot when the answers change?
Edit the source page, then refresh it. On SiteGPT, Growth refreshes websites monthly and Scale refreshes weekly. Starter, uploaded files, and connected apps need a manual resync.
Can an FAQ chatbot hand a conversation to a person?
Yes, on most platforms. SiteGPT includes human support on every plan. In a scripted flow, add a branch that passes the visitor to an agent or collects an email address.
Sources
All sources were checked on October 5, 2026.
Written by
.png?table=block&id=36e7a7e9-77da-8039-8629-c68d21d9fc2b&cache=v2)