FAQ Schema Generator

Start with the bad news

If you are here to add FAQ schema so that your search result gets the expanding list of questions underneath it, that result no longer exists. Google restricted FAQ rich results to a handful of government and health sites in August 2023, and on 7 May 2026 it retired them completely. Reporting was pulled from the Search Console rich result report and the Rich Results Test the following month, and from the Search Console API in August.

This matters because almost every other FAQ schema generator still sells the feature. They will show you a mocked-up search result with a dropdown on it. That picture has been out of date for three years for most sites and is now out of date for all of them. No markup you add, and no tool you use, will bring it back.

Now the part that is still true

FAQPage was not deprecated. Google retired one feature that consumed it; schema.org, a separate body with separate aims, still publishes the vocabulary and plenty of things still read it.

• AI answers and assistants. This is the real argument now. A growing share of questions are answered by something that read your page rather than linked to it, and clean structured question-and-answer pairs are the easiest possible thing for such a system to lift accurately.
• Other search engines. Bing and others have made no equivalent announcement.
• Your own stack. Site search, internal tooling and anything else that parses your pages gets a clean machine-readable version of the questions for free.
• It is simply accurate. Structured data describes what a page is. A page with an FAQ on it is an FAQPage, and Google has said plainly that structured data it does not use causes no problems.

What it will not do is improve your ranking. It never did.

The one way this markup can hurt you

Structured data has to describe what a visitor can actually see. Markup containing questions and answers that are not on the page, or that say something materially different from what is on the page, is a structured data violation and can earn a manual action. This is not theoretical, and it is the single most common way FAQ markup goes wrong: somebody writes the questions in the schema generator, pastes the output, and never adds the questions to the page.

That is why this tool offers to read the questions off your page instead. Give it an address, it finds the headings that read as questions and takes the text underneath each one as its answer, and the markup matches the page by construction rather than by you remembering to check.

What valid FAQPage markup looks like

The shape is small and the rules are few. A FAQPage has a mainEntity array. Each entry is a Question with a name. Each question has an acceptedAnswer of type Answer with a text. That is the whole specification.

• Use acceptedAnswer, not suggestedAnswer. The second is for pages where the public answers each other, which is QAPage, not this.
• Never leave an answer empty. An empty or missing text invalidates the entry.
• Do not repeat a question. Each has to be distinct.
• One FAQPage per page. If your SEO plugin already outputs one, use that rather than adding a second.
• Limited HTML in answers. Only h1 to h6, br, ol, ul, li, a, p, div, b, strong, i and em survive. Everything else is stripped, and it all has to be escaped properly inside the JSON string, which is where hand-written markup usually breaks.

JSON-LD in a script tag is the only format worth using. It goes in the head or the body, and Google treats both the same. Microdata and RDFa still work and mean editing your HTML in several places for no gain.

What this tool checks

Valid JSON is the easy part and the only thing most checkers do. This also reports, question by question:

• a question with no answer, or an answer with no question
• the same question twice
• a question that does not read as one, which is usually a heading that wandered into the list
• an answer too short to be answering anything
• HTML tags that Google strips, named individually
• a single-question FAQ, which is valid and almost always wanted to be a paragraph

Paste markup you already have, including a whole @graph with half a dozen types in it, and the FAQPage is picked out, read back into the builder, checked and handed to you corrected.

How many questions, and which ones

As many as the page genuinely answers. There is no limit and no target. The behaviour that got the feature killed was identical FAQ blocks stamped across hundreds of pages to farm extra space in the results, so the thing to avoid is obvious enough: write the FAQ because readers ask those questions, then mark it up.

Which tool do I want?

• This one turns a question and answer section into valid FAQPage markup, or checks the markup you already have.
• The SERP Snippet Preview Tool shows what your result will actually look like now, which is the plain title, URL and description, measured in pixels.
• The Meta Description Checker finds the pages whose descriptions are missing, duplicated or too long to survive.
• The Heading Structure Checker is worth running first if the questions on your page are headings, because this tool finds them by reading that outline.
• The Free Readability Checker and the Duplicate Word Checker deal with the writing inside the answers.

Is anything stored?

No. Questions you type and markup you paste are handled entirely in your browser, which is why the output appears as you type. The one exception is reading questions off a page by address: a browser is not permitted to fetch another site’s pages directly, so the address is sent to our server, which reads the page and returns its headings and text. Nothing is written to a database or a log either way.