In practice · 9 min read

An FAQ page AI can quote

Most FAQ pages are a bin of loose questions nobody can do anything with. This is how to write answers that help a customer and that an AI system can adopt as they are.

By NXEO · 1 Oct 2026
In short
  • A good answer starts with the answer. The first sentence has to be able to stand alone.
  • Do not hide answers so that they are only loaded after a click. The text has to be in the page.
  • Put questions with the topic they are about, and not only on one collection page.

Almost every website has a page with frequently asked questions. Almost none of them is good. It holds twelve questions nobody asks, with answers like "Please contact us for more information."

That is a pity, because a block of questions and answers is the form closest to an AI answer. Someone asks a question, the system looks for a page with that question and an answer on it. An FAQ page AI can quote is therefore one of the most rewarding things you can make.

This article describes how to do that. With examples of what works and what does not.

Why this form works so well

In essence an AI system does two things. It looks for text that fits a question, and it takes from that text a piece it can use.

A question as a heading with the answer directly beneath it makes both easy. The heading resembles what the user asked. The paragraph beneath it is the piece to quote. The system does not have to guess where the answer begins and ends.

The same goes for the AI Overviews in Google. For a question, Google runs several searches on parts of it. A page with ten clear questions and answers has ten chances of being the best answer to one part. More on that in getting into Google AI Overviews.

And it works for people. A customer with a question scans the headings, finds theirs and has an answer within ten seconds.

Rule 1: use the questions that are really asked

A frequently asked question is a question that is asked frequently. That sounds obvious, but most pages contain questions the company likes to answer. "Why should I choose you?" is not a customer's question. It is a sales pitch with a question mark.

Real questions come from your phone, your email and your first appointments. How to collect them is covered in what questions do customers ask AI.

Write them down the way the customer asks them. "What does it cost?" and not "What is the pricing policy?" A system looks for similarity with the user's question. The closer your heading is to it, the better.

Rule 2: start with the answer

This is the most important rule and the one broken most often.

The first sentence of your answer has to be the answer. Not the run up, not the context, not "that is a good question". Whoever reads only that one sentence has to know where they stand.

An example. The question is: "How long does fitting a kitchen take?"

Not like this: "The time needed to fit a kitchen depends on a large number of factors. Every situation is different and we always look at your personal wishes. We will gladly discuss the options in a meeting without obligation."

Like this: "Fitting a kitchen usually takes two to four working days. A straight kitchen without building work is in place in two days. If pipes have to be moved or walls adjusted, allow a week."

The second answer gives a number, a range and the two things it depends on. That is what a customer wants to know and what a system can adopt.

Rule 3: make every answer stand alone

An AI system takes one answer out of your page and puts it into a text of its own. The rest of your page does not come along. So the answer has to be right without what is above or below it.

That means:

  • Repeat the subject. Write "A heat pump lasts fifteen years on average" and not "It lasts fifteen years on average".
  • Do not refer to other answers. "See question 3" does not work when question 3 is not there.
  • Name your company or your town where it matters. "We work in Zwolle, Kampen and Hattem" in the answer to the question about the service area.

A good test: read the answer aloud to someone who has not heard the question. Do they understand what it is about?

Rule 4: be concrete, even when it depends

"That depends" is the honest answer to most questions. It is just not an answer. The art is to say what it depends on and what the extremes are.

Take the price question. You cannot name a fixed price, everyone understands that. You can say: "Renovating a bathroom costs between eight and twenty thousand euros with us. The difference is mainly in the size, the choice of tiles and fittings, and whether the pipework has to be replaced."

That is concrete and honest. The customer knows what to think of. And a system that is asked what a bathroom costs has a usable sentence with your name attached.

Dare to say no as well. "No, we do not fit kitchens bought elsewhere." A clear no prevents wrong enquiries and is just as usable for a system as a yes.

Rule 5: keep it short, but not too short

A good answer is between forty and a hundred and twenty words. Shorter and there is too little to work with. Longer and it is no longer an answer, but an article.

The structure that nearly always works:

  1. The answer in one or two sentences.
  2. The nuance: what it depends on, or the main exception.
  3. The next step, if there is one. A link to a page with more explanation.

If you have more to say, write a separate page about the topic and link to it. The frequently asked question remains the short answer.

Rule 6: make sure the text is really in the page

This is a technical point, and it often goes wrong without anyone noticing.

Many websites show questions as a row you can click open. That is fine, as long as the answers are already in the page and are only hidden. It goes wrong when the answer is only fetched at the moment someone clicks. A system does not click. It then sees only the questions.

You check it like this. Open the page, click nothing open, select everything and copy. Paste it into an empty document. Are the answers there? Then it is fine. Are there only questions? Then you have a problem to discuss with your website builder.

On our own frequently asked questions page all answers are in the page, including the collapsed ones.

Rule 7: put questions with the topic

One large page with fifty questions about everything is a page without a subject to a system. It does not know what it is about.

It works better to place questions where they belong:

  • On the page of a service: the five questions that are about that service.
  • On the pricing page: the questions about cost, paying and cancelling.
  • On the contact page: questions about availability and response time.
  • On a collection page: everything together, sorted by topic.

That way every page gets more content about its own subject, and the answer sits where the customer is at that moment.

May the same answer then be in two places? To a limited extent that is not a problem. It becomes a problem when you copy whole pages.

Rule 8: add markup that systems understand

There is a standard way to tell search engines that a piece of text is a question with an answer. It is described by schema.org. It is a piece of code a visitor does not see.

Is it needed? Google says you do not have to add special markup for the AI answers. It does help systems to see without doubt what the question is and what the answer is. The condition is that the code matches exactly the text that is visibly on the page.

Our advice: add it if your website system makes that easy. Many systems have a component for it. Do not expect miracles. A poor answer with tidy code remains a poor answer.

Rule 9: keep it current

A page with questions ages quickly. Prices rise, schemes change, you stop offering a service. An answer that is no longer right is worse than no answer. A system adopts it and a customer calls with an expectation you cannot meet.

Put a moment in your calendar each quarter. Go through all the answers. Are the amounts, the terms and the service area still correct? Have new questions come in on the phone?

What happens when old information stays up is covered in AI says something wrong about my company.

Five questions nearly every company has to answer

Not sure where to start? These five come back in almost every sector.

  1. What does it cost? A range with an explanation of what it depends on.
  2. How long does it take? From first contact to completion, in days or weeks.
  3. Where do you work? Towns by name, and where you do not go.
  4. What happens after I get in touch? The steps in order.
  5. What if I am not satisfied? Guarantee, aftercare or how you resolve complaints.

Whoever answers only these five well is ahead of most competitors.

Improving an answer, step by step

Take an answer as it often stands on a website.

Question: "Do you also work at weekends?"

First version: "We are a flexible company and like to think along with our customers. Get in touch to discuss the options."

What is wrong? There is no answer. The customer still does not know, and a system can do nothing with it.

Second version: "Yes, by arrangement we also work on Saturdays."

Better. There is an answer. But read on its own you do not know who "we" is and what the work is.

Third version: "Yes, our engineers also work on Saturdays by arrangement, at no surcharge. On Sundays we only work for emergencies, such as a leak. A surcharge of fifty percent on the hourly rate applies to that."

This is an answer that is right, can be read on its own and covers three follow up questions at once. The details in this example are made up. The structure is what counts.

Words that weaken an answer

Some habits creep into nearly every first draft. Removing them makes an answer shorter and stronger at once.

Hedging without content. "In principle", "generally speaking", "in many cases". If there is an exception, name it. If there is none, drop the hedge.

Praise for yourself. "Our experienced specialists", "our high quality service". A customer skips it and a system cannot quote it. Say what the specialist does.

The promise instead of the fact. "We always aim to respond quickly" says less than "we reply within one working day".

The deferral. "Feel free to ask us for the possibilities." That sends the customer away with the question still open. Give the answer, and offer contact afterwards for the details.

Read your answers once with only this in mind. What remains after the cutting is usually the answer.

What it delivers

A good set of questions and answers does three things at once.

You get fewer phone calls with the same question. You get enquiries from people who know better what they are starting. And you give AI systems exactly the pieces of text they need to mention you.

It is also the work you can start on fastest. Ten questions, ten answers of sixty words. That is an afternoon.

And after that?

Ten questions is a start. After that come the questions per service, the questions that are added, and the upkeep. That is the part that is left undone at most companies.

It is also part of what we do. We measure for which questions you are missing, write the answers in your tone and publish them in the right place in your website. Curious which questions you are missing now? Request the free baseline check, or read what NXEO is.

Curious how AI mentions your company today?

Enter your website. Within 48 hours the report is in your inbox, free.

Read next

Where can we send the report?

In your inbox within 48 hours. No newsletter, no spam.

Website