Schema Markup for Answer Engine Optimization: What if Does & What it Doesn't

Google's own documentation says this out loud: "Structured data isn't required for generative AI search, and there's no special schema.org markup you need to add."

That sentence kills most of the advice currently ranking for this topic. It does not kill schema. The same page goes on to say it's still worth using, "as it helps with being eligible for rich results on Google Search."

So schema is not a cheat code for AI citations. It's still worth doing. Here's which types to use, how to write them, and what you can honestly expect.

What is Schema Markup?

Schema markup is code that labels the information on a page. What the page is, who wrote it, who published it, when it was updated, what it connects to.

Two words get used interchangeably and shouldn't be. Schema.org is the vocabulary, the shared dictionary of types like Article and Organization and the properties that belong to them. JSON-LD is one of three formats for writing that vocabulary into a page, alongside Microdata and RDFa.

Google supports all three and says any of them is fine when the markup is valid. It also recommends picking "a format that's easiest for you to implement and maintain (in most cases, that's JSON-LD)." Take the recommendation. JSON-LD sits in one script block, stays out of your HTML, and one person can own it.

At PBJ Marketing, we tend to favor JSON-LD.

Does Schema get you into AI answers?

Three different things get collapsed into one question here. Pull them apart and the answer gets clearer.

Worth separating one more thing: being cited in an AI answer is not the same as being in a model's training data. Citation happens at query time, when a system retrieves sources to ground an answer. Training happens on a completely different schedule. Schema controls neither.

The only published test on this is worth reading: Otterly.ai deployed five schema types on their own site from December 2025 to March 2026, tracking brand coverage across seven AI platforms with 319 prompts and using competitor brands as a control. 

• Six of the seven could not fetch or interpret the schema when asked directly. Gemini was the only one that pulled the JSON-LD correctly. 
• Google AI Mode invented a schema type that wasn't on the page. 
• Their AI Overviews coverage did climb over the window, but so did the control brands' coverage, which points at an algorithm change rather than anything the markup did.

It's the only real evidence anyone has published, and it points somewhere useful: schema most likely feeds the index that AI answers retrieve from. It probably isn't being read off your page by a model.

Which Types to Prioritize

Pick types that describe your content accurately. That's the whole rule. There is no universal AEO stack, and anyone selling you one is selling you a checklist.

If you run a services business, the most useful type on that list has nothing to do with content. It's Organization with sameAs, pointing at your verified profiles. 

That's entity disambiguation, and it's the strongest honest argument for schema in an AI context. You're telling every system that reads your site which company you are and where else you exist.

What the Markup Looks Like

Google's Article documentation has no required properties. Everything is recommended, which means you add what applies and skip what doesn't. Here's a working BlogPosting block:

json
{
  "@context": "https://schema.org",
  "@type": "BlogPosting",
  "headline": "Schema Markup for Answer Engine Optimization",
  "author": {
    "@type": "Person",
    "name": "Jane Doe",
    "url": "https://example.com/team/jane-doe/",
    "sameAs": ["https://www.linkedin.com/in/janedoe/"]
  },
  "publisher": {
    "@type": "Organization",
    "name": "Example Co",
    "url": "https://example.com/"
  },
  "image": "https://example.com/img/schema-aeo.jpg",
  "datePublished": "2026-09-29T09:00:00-04:00",
  "dateModified": "2026-09-29T09:00:00-04:00"
}

Four rules that matter more than the code itself:

• Your headline should match the H1 a reader sees. Google is explicit that you shouldn't mark up information that isn't visible on the page.
• Every author credited on the page belongs in the markup, each in their own author field. Don't jam two names into one string.
• dateModified moves when the content changes. Not when you push CSS. A date that updates on every deploy is noise.
• Your publisher block should be byte-identical across every page on the site.

Before you write any of this, view source and search for application/ld+json. Most WordPress sites are already emitting schema from a plugin, sometimes from two plugins at once, and those blocks fight each other. Fix what's there before you add more.

FAQ Schema is Not What it Was

Google added a deprecation notice to its FAQPage documentation on May 8, 2026, saying the feature would stop appearing in Search on May 7, 2026. The documentation came down on June 15, 2026.

You will still find plenty of posts telling you to add FAQPage markup for rich results. Most of them were written before the change and nobody went back. Date-check any schema advice before you act on it, including this post.

Keep your FAQ section. Questions and short answers are one of the most extractable formats for AI systems, and readers use them. Just stop treating the markup as a rich-result play.

How to Test Schema

Two tools, two jobs, and people confuse them constantly.

• The Schema.org Validator checks whether your markup is valid Schema.org. 
• Google's Rich Results Test checks whether you qualify for a Google search feature. Valid markup that Google doesn't support will pass one and fail the other, and that's correct behavior, not a bug.

The failures we see most: two plugins emitting conflicting blocks, an author field pointing at an admin account, dates that never move, markup describing content that isn't on the page, and business details that have drifted out of sync with the site and the business profile.

Passing validation confirms your syntax but it doesn't promise you anything else.

How to Measure Schema Markup

Track four things and keep them apart: organic performance, AI impressions or citations, referral traffic from AI surfaces, and conversions.

• Search Console's Generative AI performance report covers AI Overviews and AI Mode. The metric is impressions, broken out by page, country, date, device, and search type. Not clicks. Not position. It skips Search Labs experiments, needs a minimum volume to show anything, and is still rolling out, so your property may not have it yet.
• Bing Webmaster Tools' AI Performance report, in public preview since February 2026, reports citations across Copilot, Bing's AI summaries, and some partner integrations. Total citations, average cited pages, grounding queries, page-level activity.
• An impression is not a click, and a citation is not a visit. Report them as what they are.

Now the hard part: proving the schema did anything. Set a baseline before you touch the page, then change only the markup and wait a few weeks. If you also rewrote the intro, republished the post, and earned three links that month, you have no way to separate schema from everything else you did.

Even a careful test can miss this. Otterly watched their AI Overviews coverage climb after adding schema. Then they checked the competitor brands they were tracking as a control and found the same climb. The lift came from an algorithm change, not the markup.

Where to Start

• Audit what your CMS already outputs, fix the conflicts, add Organization with sameAs, then Person for every author, then BlogPosting on your content, then BreadcrumbList. 
• Validate in both tools. 
• Set a baseline in Search Console and Bing before you touch anything else.

That's a week of work. Do it because clean structured data makes your site easier for every system to read. Don't do it because someone promised you an AI citation.

Want a second set of eyes on your structured data? Our SEO team audits it as part of every technical engagement.

‍