Most FAQ pages are a graveyard. Someone dumped twelve questions nobody asks at the bottom of a template, marked it up with schema, and moved on. That page does nothing for search and less for AI. A good FAQ page is the opposite: a tight set of real questions, answered in language a model can lift cleanly, placed where the answer actually belongs.
The format matters now in a way it didn’t five years ago. Search engines stopped showing FAQ rich results for most sites, so the old reason to build these pages is mostly gone. The new reason is bigger. AI assistants pull answers from pages that read like answers, and a well-built FAQ is the cleanest answer format you can publish.
Why FAQ pages still earn their place
Google retired FAQ rich results for the vast majority of sites in 2023. The accordion stars and expandable questions that used to widen your listing are now reserved for a narrow set of authoritative government and health domains. If your only goal was the rich result, the FAQ page lost its job.
But the underlying behavior shifted somewhere more useful. People ask questions in full sentences now, to Google’s AI Overviews, to ChatGPT, to Perplexity, to Gemini. Those systems reward content that states a question and answers it in a self-contained chunk. An FAQ entry is exactly that shape: a clear question, a complete answer, no dependency on the paragraph above it.
A search engine ranks your page. An answer engine quotes one sentence from it. FAQ pages are built for the second job.
So the FAQ page earns its place for two reasons. It captures the long-tail, conversational queries that map almost word for word to a question heading. And it gives AI systems pre-chunked, quotable units that survive being pulled out of context, which is how AI assistants cite sources in the first place.
Find the questions worth answering
The fastest way to ruin an FAQ page is to invent the questions. Real users ask specific, awkward, sometimes repetitive things. Your job is to find those, not to brainstorm what sounds reasonable.
Start with sources that record actual phrasing:
- Your sales and support inbox. The questions people ask before they buy, in their own words, are the highest-value entries you will ever write. Read the last two months.
- Search autocomplete and “People also ask.” Type your core topic into Google and harvest the suggested completions and the related questions box. These are real queries, ranked by frequency.
- The AI assistants themselves. Ask ChatGPT and Perplexity the question a customer would ask about your category. Note what they answer, where they hedge, and what they get wrong. The gaps are your openings.
- On-site search logs. What people type into your own search box, especially the queries that return nothing, tells you what they expected to find and didn’t.
Group what you collect by intent. Pricing questions, comparison questions, how-it-works questions, and objection questions each behave differently and sometimes belong on different pages. Treat the page as keyword research expressed as conversation: the same demand, captured in the phrasing people actually use.
One filter cuts the list down fast. If the question has a real, non-obvious answer that helps a buyer decide, keep it. If the answer is “yes” or “it depends” with nothing behind it, drop it. Padding dilutes the page and trains models to ignore it.
Write the answers so machines can lift them
This is where most FAQ pages fail. The question is fine. The answer rambles, buries the point, or assumes you read the three entries above it. AI systems extract in chunks, so each answer has to stand completely on its own.
A few rules that hold up:
Answer in the first sentence. Lead with the direct answer, then add the nuance. “An FAQ page helps with AI visibility because it formats content as self-contained question-and-answer pairs that assistants can quote without rewriting.” The first sentence is the part that gets pulled. Make it true and complete on its own.
Repeat the question’s key terms in the answer. Don’t write “Yes, it does.” Write “Yes, FAQ schema is still worth adding,” so the answer survives separation from its heading. Models and snippets often show the answer without the question attached.
Keep each answer between 40 and 80 words for most entries. Long enough to be complete, short enough to quote whole. If an answer needs three paragraphs, it is not an FAQ entry. It is a blog post, and you should write that post and link to it.
Use real question headings. Mark each question as an actual heading element (h2 or h3), phrased the way a person would ask it. “How much does X cost?” beats “Pricing.” The heading is a ranking and extraction signal, and it mirrors the spoken query.
Write one fact per sentence where you can. Dense, comma-spliced sentences are hard to extract cleanly. Short declaratives give an assistant a clean unit to lift. It is the same discipline that makes any page legible: structure a reader and a parser both follow.
Mark it up, but know what the schema does now
You should still add FAQPage structured data, with one honest caveat: it will almost certainly not produce a visible rich result for you anymore. What it does instead is state, in machine-readable form, that this block is a question and this block is its answer. That removes ambiguity for the systems parsing your page, and ambiguity is what costs you a citation.
The markup is straightforward JSON-LD. A FAQPage contains a list of Question items, each with an acceptedAnswer. The full mechanics, plus the validation steps and the mistakes that quietly break it, live in our guide to structured data for brands, so I won’t duplicate the code here.
Two rules keep you out of trouble. The schema must match the visible text on the page exactly, word for word. Marking up answers a user can’t see is a guidelines violation and a fast way to get the markup ignored. And use FAQPage only for genuine question-and-answer content. Do not wrap a product description or a marketing pitch in question shapes to game the format. The systems reading it are good at spotting that, and the penalty is being filtered out.
Where the FAQ belongs
The standalone “/faq” page is rarely the best home for these answers. Questions answer buying decisions, so they work hardest next to the decision.
Put pricing questions on the pricing page. Put product objections on the product page. Put implementation questions in the docs. A short, contextual FAQ block at the bottom of a service or product page outperforms a giant central FAQ page that nobody scrolls, because it answers the exact doubt a reader has at the exact moment they have it.
Keep a central FAQ page too, if your topic is broad enough to warrant one. But treat it as a hub that links out, not a dumping ground. The contextual blocks do the converting. The hub catches the stray search. This is part of broader generative engine optimization: putting answers where both people and models expect to find them, in the structure they can read.
How Strynal approaches FAQ pages for AI and SEO
We don’t bolt an FAQ page on at the end of a build. We treat the questions as content with a job, mined from real sales conversations and search behavior, written to be quoted, and placed against the decision they unblock. The schema is the last step, not the strategy.
That work sits inside how we think about AI visibility: make the brand legible to the systems people now ask for recommendations, by structuring real expertise instead of chasing tricks that get patched out. A clean FAQ is one of the most durable pieces of that, because it helps a human and a machine in the same move.
If your FAQ page is a graveyard, or you don’t have one and the questions are piling up in your inbox, start a conversation. We’ll figure out which questions are worth answering and where each one belongs.