FAQ pages still work in 2026 (if you build them right)

In 2023, Google deprecated rich-result support for FAQPage markup on most domains. A wave of "FAQ schema is dead" content followed. Many SaaS teams ripped out their FAQ blocks or stopped maintaining the schema. Three years later, the picture is different than that take suggested.

FAQ pages didn't stop working. They stopped winning Google rich results, which is a different claim. For AI answer engines they're actually more valuable than they used to be, with one specific condition.

#What AI engines do with FAQPage markup

When an LLM is searching for an answer to a user's question, it strongly prefers content where:

• The question is stated explicitly
• The answer is short and self-contained
• The pairing is unambiguous

FAQPage schema delivers exactly this. Each entry is a {question, answer} pair, machine-readable, atomic. When the engine is doing retrieval against a query, FAQ-marked content matches well. When it's lifting an answer, FAQ content is cheap to quote.

In our experience, pages with proper FAQPage schema tend to be cited somewhat more often than equivalent pages without it.

That's a smaller lift than Article schema, which tends to help more, but it's free and additive.

#The condition

The condition is "your FAQ has to actually be a FAQ."

The pattern Google penalized in 2023 (and AI engines learned to penalize too) is the marketing FAQ. The block at the bottom of a landing page where the "questions" are softballs ("Why is your product the best?") with answers that are pure sales copy. That pattern not only doesn't help, it hurts. Engines learned the difference between editorial FAQ and "fake-Q-marketing-A."

What works:

• Pages whose primary content is genuine question and answer (support docs, help center articles, troubleshooting guides)
• FAQ sections at the bottom of feature pages that surface real, specific questions ("Does this integrate with Salesforce?" not "Is this the best CRM ever?")
• Documentation Q-and-A from real support tickets, anonymized

What doesn't work:

• Three softball "questions" added to every product page to game the markup
• "FAQs" generated by SEO tooling that don't reflect actual customer questions
• Boilerplate "Why choose us?" framings

#How to write good FAQ content for AI

A few patterns we see consistently in cited FAQ content:

#1. Questions are how a user would actually phrase them

Real customer language. "Can I export my data?" not "What data export functionality is available?"

If your support team has a stash of common questions, that's the source material. If they don't, you have a separate problem worth fixing.

#2. Answers lead with a verdict, then explain

Bad: "There are several considerations to weigh when exporting data. Depending on your plan, you may have access to..."

Good: "Yes. Every paid plan includes CSV and JSON export from the account settings page. Free plans can export once per month."

The first sentence is the answer the engine will lift. Make it the answer.

#3. Length is bounded

Roughly 30 to 150 words per answer. Shorter and the answer is too thin to be useful. Longer and it stops looking like a FAQ and the engine treats it as regular article content.

#4. Cross-link to the deep page

Each FAQ answer can link to the longer treatment if there is one. This is good for users and for the engine's understanding of your site graph.

#The minimal FAQPage schema block

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "Can I export my data?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Yes. Every paid plan includes CSV and JSON export from the account settings page. Free plans can export once per month."
      }
    },
    {
      "@type": "Question",
      "name": "Does this work with Salesforce?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "Yes, via our native Salesforce integration available on Pro and above. Setup takes about 10 minutes and is documented in our integration guide."
      }
    }
  ]
}

#Where to put it

Three places that work well:

1. Help center articles. Each article is a FAQPage with one or more Q-A pairs. This is the strongest use case.
2. Product feature pages. A small FAQ section at the bottom addressing the three or four real questions buyers ask. Mark only the genuine ones.
3. Pricing page. "Can I change plans later?" "Do you offer annual billing?" These are real, specific, and lift well.

Where NOT to put it:

• Homepage (too generic, low signal)
• Press release pages
• Blog posts (use Article schema instead, with FAQPage ADDITIONAL only if the post genuinely contains a Q-and-A section)

#The summary

The FAQ schema lift is smaller than you might want, but it's real, free, and stacks with the rest of your structured-data work. The critical thing is that it must reflect real Q-and-A content. The old "stuff FAQ markup on every page" pattern is detected and penalized. Use it where you have genuine question-and-answer material, skip it everywhere else.