FAQPage Schema After Google Dropped the Rich Result: Keep It or Cut It?

Short answer: Keep FAQPage markup on any page that has a visible list of questions and answers, and cut it everywhere else. Google removed the expandable FAQ rich result from search for all but government and health sites in August 2023, so the markup no longer changes how the page looks in results. It still validates, it still describes the page accurately to any parser, and it still ties each question to its 40-to-60-word answer in a form that is trivial to extract. What it does not do is earn a citation by itself. The answer earns the citation; the markup labels it.

What happened in 2023?

In August 2023 Google announced it would show FAQ rich results only for "well-known, authoritative government and health websites", and HowTo rich results not at all on desktop. For law firm sites that meant the accordion of questions under a result vanished. Agencies that had added FAQPage to every page for the visual effect stopped, and some removed it.

The reasoning for removing it was that it no longer did anything visible. The reasoning for keeping it is that "visible in Google's results" was never the only reader.

Who reads it now?

• Google's own systems, which still parse it and use it to understand the page even without a rich result.
• Bing, which never dropped its FAQ treatment entirely and whose index feeds Copilot and, in part, ChatGPT search.
• Anything that extracts question-answer pairs from structured data, which is the cheapest way for a system to get a clean answer to a clean question.

Whether the AI answer engines weight it specifically is not something anyone outside those companies can show. In our own logs, pages with and without the markup are cited at similar rates when the visible answers are equally good, which is consistent with "the answer matters, the label does not hurt".

What is the rule?

Keep it when all of these are true:

1. The questions and answers are visible on the page, in full, as text.
2. Each answer is complete and specific (the 40-to-60-word answer).
3. The questions are ones a client would ask, not keyword variants of each other.
4. The markup matches the page exactly: same questions, same answers, same order.

Cut it when any of these is true:

• The questions are in a collapsed accordion whose content is not in the HTML.
• The markup lists questions the page does not show, or shows different answers.
• The "FAQ" is one question added to satisfy a checklist.
• The page is a product or service page where the questions are promotional ("Why choose Smith & Lee?").

Google's guidelines require that the marked-up content be visible to the user and not used for advertising; the second list is what gets structured data ignored or, in a manual review, penalized.

What is the JSON?

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "How long does probate take in Los Angeles County?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "A straightforward probate in Los Angeles County takes nine to eighteen months from filing to final distribution. The minimum is set by the four-month creditor claim period after Letters issue; the rest is hearing availability at Stanley Mosk and how quickly the inventory and accounting are completed."
      }
    },
    {
      "@type": "Question",
      "name": "Does a primary residence under $750,000 still need probate?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "For deaths on or after April 1, 2025, a primary residence valued up to $750,000 can pass by a petition to determine succession under AB 2016 instead of a full probate. Other real property, and residences above the limit, still go through probate unless held in a trust."
      }
    }
  ]
}

One FAQPage per page, five to ten questions, each answer plain text without links or markup. The answers are the same text as on the page; do not write a shorter version for the JSON.

Where should it not be combined?

Do not put FAQPage on a page that is also marked Article unless the FAQ is the article; use one main type per page and attach the FAQ as a separate node in the same graph only when the page genuinely has both. Do not nest it inside LegalService. Keep the graph simple: organization, page, person, and, where it applies, FAQ.

What does it do for AI citations?

Nothing on its own. The pages we see cited for question-shaped queries are cited because the visible answer is the best one available, and most of them happen to carry the markup because we add it as a matter of course. Adding it to a page with vague answers does nothing. Adding it to a page with precise answers labels what was already going to be quoted.

How do you check the site?

Run each page with a FAQ through the Schema Markup Validator; confirm every marked-up question appears verbatim on the page; remove the markup from pages that fail. Our schema generator produces the block from the visible questions, and the law firm schema guide covers how it fits with the rest of the graph. The free visibility check reports which pages carry structured data that does not match the page, which is the version of this that causes problems.