You write FAQs by starting from the real questions your customers ask, in their own words, and giving clear, self-contained answers that put the substance in the first sentence (about 40–60 words as a practical guide). A few real questions with substantive answers are worth more than many generic ones. FAQPage schema is supporting only — it doesn't guarantee appearance in AI answers; the value is in the visible content.
When someone asks ChatGPT "which accounting firm handles startups in Thessaloniki," they don't just type two or three keywords. They phrase a full question, the way they'd ask a person.
When an AI tool uses search or retrieval, it can find pages that answer the user's specific question. That's why a well-written FAQ section can be a useful part of your website.
It isn't an SEO trick. Its structure — a specific question and a clear answer — presents the information in a way that's easy for the reader to understand and potentially easier for a search or retrieval system to use. That, however, doesn't guarantee the page will be picked as a source.
Many FAQs, though, aren't written that way. They include questions no one actually asks, with answers so vague they end up helping neither the reader nor a search or AI system.
Let's look at how to write an FAQ section that has real value for the customer and stays clear for search and AI systems.
Why do FAQs matter for AEO?
The question-and-answer format fits the way people use many AI tools.
Questions put to an AI tool are often more detailed and phrased more naturally than the short queries we type into Google. An FAQ written in the customer's words sits closer to how they'll actually phrase their question.
There's a second, more technical reason.
Search and retrieval systems can draw on specific passages of a page rather than necessarily the whole text as a single unit.
A self-contained question with a complete answer creates a clearly organised piece of content for the reader. Search and retrieval systems can draw on specific information from a page, but there's no public documentation showing that FAQ format, on its own, receives preferential treatment from AI engines.
For Google specifically, there's no requirement to break content into small chunks or write in a special way for its generative AI features. A well-structured regular section, with a clear heading and a direct answer, can be just as useful.
Take a simple example.
Someone asks:
"What time is check-out?"
If the information sits inside a long paragraph with all the accommodation terms, it is less immediately visible and has to be separated from the surrounding text.
An FAQ section presents it clearly:
"Check-out is until 12:00. If there's availability, we can extend it to 14:00 at no charge, or later for the cost of half a night."
You didn't change what the system can do. You simply presented the information in a clearer, easier-to-find form.
Which questions do you include?
You include the questions your customers actually ask, phrased in their own words. Not only the ones the business finds convenient to answer.
The value of the whole section depends largely on that choice.
The question:
"What are our opening hours?"
is written from the business's side.
The customer is more likely to ask:
"Are you open after 6 p.m.?"
The right questions already exist. You don't need to invent them; you need to write them down.
You can find them:
- in the emails customers send before an appointment or a purchase,
- in the questions your team gets on the phone every day,
- in conversations with your sales or front-desk staff,
- in customer reviews,
- in your Search Console data, related searches and Google's "People Also Ask,"
- in suggestions generated by AI tools, only as a starting point for research and not as proof that those specific questions are actually asked by customers.
Suggestions from an AI tool should be checked against real data from searches, sales, customer service, or audience research before they are turned into content.
Pay attention not only to the topic of the question, but also to its wording.
The phrase:
"What is your cancellation policy?"
is the language of the business.
The phrase:
"Until when can I cancel free of charge?"
is closer to how the customer will ask.
A good test is to read the question on its own. It should be complete and understandable even if the reader doesn't know which page they're on.
This:
"Do I need an appointment, or can I just come?"
is a complete question.
This:
"Information"
isn't a question and doesn't explain what the customer wants to know.
How long should an answer be?
For most questions, a short paragraph is a good starting point.
40 to 60 words can serve as a practical editorial guide, but not as a technical requirement of AI tools. In some cases a single sentence is enough, while other questions need more explanation. The answer should be as short as the topic allows and as complete as it needs to be.
The basic principle stays the same:
The answer first, then the details.
Many businesses do the opposite. They start with a general introduction and leave the substantive answer for the end.
That creates a problem, because it delays the substantive information and makes it harder for the reader to find the answer quickly. Putting the answer up front also makes the relationship between the question and the answer clearer, without creating a special technical advantage or guaranteeing selection by an AI tool.
Both extremes fail.
A two-word answer, such as:
"Yes, we can."
explains nothing when read without the question above it.
On the other hand, a three-hundred-word answer is usually longer than an FAQ needs and can make it hard for the reader to find the substantive information quickly.
What you need is a clear answer that stays understandable even without the rest of the page.
How do you write an answer that stands on its own?
You put the substantive answer in the first sentence, keep one idea per question, and write in everyday language.
The first sentence should answer fully, as if it's the only part the user will read. What follows should reinforce the answer, not delay it.
Look at the difference between two answers to the same question.
❌ "The matter of delivery time depends on many factors and varies case by case, but generally we try to be as fast as possible."
✔ "Orders ship within 1–2 business days and arrive in 2–4 days on the Greek mainland. For the islands, allow 1–2 extra days. If you order before 1 p.m., the order ships the same day."
The second version answers straight away and includes specific times and numbers. It stays understandable even without the rest of the page.
It's an answer that can be used as-is, with no extra explanation needed.
To keep an answer clear:
- keep one question at a time, and don't try to answer three different topics together,
- use everyday language and avoid the business's internal jargon,
- add specifics, such as hours, prices, times and conditions,
- avoid answers like "contact us for more information" when you can give a substantive one,
- when the answer depends on the case, explain clearly which factors it depends on.
Uncertainty doesn't mean you can't answer. It means you have to explain the conditions clearly.
Do you always need FAQPage schema?
No.
FAQPage is a Schema.org structured data type for web pages that present one or more frequently asked questions and answers.
It provides a structured description of the content, but there's no public documentation that it, on its own, raises the odds of a page being picked as a source by AI tools.
The greater value lies in the questions themselves and in the visible answers the visitor reads.
Two changes are worth knowing.
Since 7 May 2026, Google has stopped showing FAQ rich results in Search. On 15 June 2026 it also removed the related documentation from Search Central.
The FAQPage type still exists in Schema.org and can be used when a web page genuinely presents one or more frequently asked questions and answers. It no longer produces an FAQ rich result on Google, and there's no public documentation that it increases citations from AI tools.
Alongside that, an Ahrefs study tracked 1,885 pages that added JSON-LD and compared them with 4,000 matching control pages. In the 30 days after the change it found no meaningful increase in citations in Google AI Mode or ChatGPT, while in Google AI Overviews it recorded a small but statistically significant drop relative to the matched control pages, without the data allowing us to conclude that schema caused it.
One important clarification is needed, though: the research looked at different schema types together, not FAQPage specifically. It also covered pages that already had a significant presence in AI citations.
So it doesn't prove schema offers no technical value. It shows that, for the pages and platforms examined, adding JSON-LD on its own didn't meaningfully raise citations.
The conclusion is simple.
Write the questions and answers well first, and make sure they appear clearly on the page.
Then you can use FAQPage when the page itself genuinely presents one or more frequently asked questions and answers. The structured data must describe the visible content accurately and must not add different or hidden information.
Schema is a supporting technical element, not a prerequisite or guarantee of appearing in AI answers.
FAQPage schema should describe only what actually appears on the page.
When does an FAQ section hurt instead of help?
When it fills the page with questions no one asks, just so there's an FAQ section.
Ten general questions with vague answers don't raise the odds of your page being used as a source. On the contrary, they weaken the content and tire the reader.
Three real questions with clear, substantive answers are worth more than ten questions added just to fill the page.
You also don't need FAQs on every page.
A section like this makes sense when the customer has real questions before deciding. For example:
- on a service page,
- on a product page,
- on a pricing page,
- on a booking page,
- in an article that covers a topic in full.
On a page where the visitor has no real questions, or isn't asked to take any action, an FAQ section usually adds no real value.
The section's position on the page matters too.
On commercial pages — such as service, product, pricing, or booking pages — it often makes sense to place it near the decision point, after you present the service or product and before the call to get in touch, buy or book. In an informational article, its position should follow the natural flow of the content.
The article's basic rule is one:
Real questions, clear answers, and content visible on the page.
The clearer and more specific your information is, the more useful it is for people and the less ambiguity it leaves when a search or retrieval system processes it. That doesn't mean AI tools use only FAQs or only information written in a particular format.
At Praion we work on FAQs together with the rest of the content, always starting from the questions your customers already ask.
AEO aims to improve a business's ability to be found, understood, and potentially appear in relevant AI answers — it doesn't guarantee a citation, recommendation, ranking, or result. AI answer systems continuously change how they select sources.