Which schema types help AI extraction after FAQ rich results ended
Use schema to clarify facts, not to chase dead SERP features

What changed when FAQ rich results ended?

The big change is simple. FAQ rich results fully deprecated and stopped appearing on 2026-05-07. That removed the main reason many teams had for spraying FAQPage markup across every template.

A lot of schema advice is still stuck in the old SERP mindset. Add more markup, win more visual treatment, call it strategy. That is not the job now. If you care about AI answers, grounding, citations and clean fact extraction, the useful question is whether the markup reduces ambiguity about who you are, what the page is for and which claims belong to which entity.

That means FAQPage moved from center stage to supporting role. It can still help you express a question and answer structure already present on the page, but it is no longer evidence that your content will be surfaced, cited or quoted. If the source page is vague, thin, rendered poorly or mixes multiple claims without clear ownership, schema does not rescue it.

Which schema types actually help AI extraction?

The short version is that helpful schema maps cleanly to page intent. You want markup that tells a machine what the page is, who published it, which primary entity it covers and how the page fits into the rest of the site.

If I had to prioritize for most B2B sites, I would start with Organization, BreadcrumbList, Article or Service depending on the template, then Person where author trust matters. Product belongs on true product pages, not because someone wants rich looking JSON-LD in every template.

This is the part people skip. Schema works best when it mirrors the visible structure of the page. If the page headline says one thing, the body drifts into another topic and the markup names a third entity, extraction gets worse, not better.

Why do these types help more than FAQPage now?

Because extraction systems need stable context more than decorative markup. They are trying to resolve entities, map relationships and lift concise facts without mangling meaning. Organization tells them who is speaking. Article tells them this is editorial material. Service or Product tells them the page has commercial intent around a specific offer. BreadcrumbList helps define where the topic sits within the site.

FAQPage is narrower. It helps only if the question and answer pairs are genuinely useful and already visible on the page. It does not solve brand inconsistency. It does not make a weak service page suddenly quotable. It does not compensate for missing author identity, thin source material or unclear page ownership.

There is another practical issue. AI crawlers do not execute JavaScript. The verified server log finding is that they fetch JavaScript files and never run them. So if your schema only appears after client side rendering, many AI crawlers will never see it in the DOM they actually process.

How should you choose schema by page type?

Use the smallest set that makes the page legible to a machine. Not the biggest set you can squeeze into a plugin.

Homepage

• Organization
• WebSite
• BreadcrumbList if visible breadcrumbs exist in the template

The homepage should establish the brand entity cleanly. Name, description, same visible positioning used across the site and no contradictory claims.

Editorial guides and research

• Article
• Person for the author when there is a real author page and consistent authorship
• BreadcrumbList

This combination helps answer systems understand who published the piece, who wrote it and where it belongs. It also reduces the common problem of useful pages being treated as anonymous content blocks.

Service pages

• Service
• Organization
• BreadcrumbList
• FAQPage only if the questions are genuinely part of the page and useful

Service schema helps if the page is about one defined offer. If it is a broad capability page covering multiple offers, forcing one Service object can oversimplify and distort the content.

Product pages

• Product
• Organization where appropriate
• BreadcrumbList

This is where attributes matter most. Keep them aligned to visible product facts. Do not borrow ecommerce patterns for agencies, consultants or software service hybrids unless the page is truly a product page.

What does this advice fail to fix?

A lot, honestly. Schema is not a substitute for extractable writing, server rendered content, crawl access or external corroboration. If assistants keep citing directories, review sites or aggregators instead of you, the issue may be source trust, not missing JSON-LD.

This advice also fails on pages that are conceptually messy. If one URL tries to rank, explain, convert and compare at the same time, adding more schema types usually increases confusion. The page needs a clearer job before it needs more markup.

It is also a poor fit for teams looking for a one line KPI. There is no honest way to say a given schema type will multiply AI mentions by some dramatic factor. A lot of those circulating GEO numbers are unsourced, and I would not build a roadmap on folklore.

And if your site depends on client side rendering for critical content, fix that first. Machines cannot extract what they do not receive in the initial response. Schema loaded late is still late.

Who should not follow this playbook?

Do not start here if your real problem is outbound execution, pipeline creation or multichannel sequencing. That belongs with the operating team on the parent brand, not this site. We run managed outbound under Outbound Pros, but execution strategy is a separate problem from AI extractability.

Also skip this playbook if your site has no stable information architecture yet. Breadcrumb and entity schema are useful only when the underlying structure is real. Marking up chaos does not create order.

Finally, this is not for teams hoping llms.txt will do the heavy lifting instead. Google states llms.txt is not used by Search, and the study across about 300000 domains found low adoption, none among the top 1000 sites and no citation lift after controls. Publish it if you want a tidy file for human readers or tool compatibility, but do not treat it as a replacement for page level clarity.

What is the practical rollout order?

• Audit which templates already have stable page purpose
• Ensure critical content and schema are present server side
• Add Organization and WebSite on the homepage
• Add BreadcrumbList where visible breadcrumbs are real and consistent
• Map Article and Person to editorial templates
• Map Service to true service pages, Product to true product pages
• Keep FAQPage only where users genuinely need question and answer structure
• Check that visible content and schema properties say the same thing

That order is boring, which is why it works. The gains usually come from consistency and reduced ambiguity, not from discovering some hidden schema object everyone else missed.

If you want a quick benchmark, ask whether a machine seeing only the rendered HTML plus schema could answer three questions cleanly: who is this, what is this page for and which specific facts belong here. If the answer is no, add clarity before you add coverage.