Schema markup that moves the needle for AI search
Crawlmind Engineering··3 min read
If you've read three schema.org guides you've read 30. The same hundred types get listed, the same generic advice about "structured data helps SEO," and you walk away no closer to knowing which 90% to ignore.
This is the short version, based on what we see across customer audits.
#The five that actually matter
In rough order of leverage:
#1. Article / BlogPosting with author and dateModified
This is the single highest-impact piece of markup for AI citation likelihood. Specifically:
{
"@type": "BlogPosting",
"headline": "...",
"datePublished": "2026-04-20",
"dateModified": "2026-04-20",
"author": {
"@type": "Person",
"name": "Sasha Lee",
"url": "https://example.com/team/sasha"
},
"publisher": {
"@type": "Organization",
"name": "Example",
"logo": { "@type": "ImageObject", "url": "..." }
}
}
The two non-negotiables are author (linked to an author page that
exists and has a bio) and dateModified. AI engines treat author
attribution as the strongest single E-E-A-T signal, and they treat
content with no dateModified as effectively undated, which for
fast-moving topics like AI tooling means "probably out of date,
deprioritize."
#2. Product with aggregateRating and offers
If you're a SaaS, your pricing page should have this. ChatGPT and
Perplexity both lean hard on Product schema when answering "how much
does X cost" queries, and the answers they generate are noticeably
more accurate when the source page has explicit price,
priceCurrency, and availability properties rather than just
rendering "$29 per month" in HTML.
The catch: aggregateRating is only legitimate if you genuinely
collect ratings. Don't fabricate. Both Google and Perplexity have
been increasingly willing to penalize sites for fake review markup.
#3. BreadcrumbList
The cheapest piece of schema to add and one of the most underrated. Breadcrumbs help AI engines understand site hierarchy, which in turn helps them decide whether a URL is a "section index" (less likely to be cited) or a "leaf article" (more likely). Sites with consistent breadcrumbs tend to be cited more readily, since the markup makes site structure explicit.
#4. FAQPage
For pages that genuinely contain Q and A content. The "atomic answer"
section of an AI response often lifts directly from a FAQPage
block. Stuffing a FAQPage into a marketing page just to get the
markup is counterproductive (Google explicitly demoted that pattern
in 2023 and AI engines learned the same lesson).
If you have a real FAQ, mark it up. If you don't, don't.
#5. Organization with sameAs links to social
This one is invisible to most SEO tools but matters for AI engines
trying to establish whether you're a real entity. sameAs pointing
to your verified LinkedIn, X, GitHub, Crunchbase, and Wikipedia (if
you have one) feeds the engines a graph of "this entity is real and
the same across these surfaces." It's the closest thing to a
machine-readable "this site is a real company."
#What to skip
A short list of types you can probably ignore unless you have a specific reason:
WebSite(a site-wide block already covers most use cases)Personprofile pages with 20 properties (engines use only a handful)Event(only relevant if you actually run events)- Most of the
CreativeWorksubtypes HowTo(Google deprecated rich-result support; AI engines still read it but the lift is modest)
#How to validate
The two tools we use:
- Schema.org validator for syntax
- Google Rich Results Test for "would this actually render"
If you're a Crawlmind customer, the audit checks all five of the high-leverage types automatically and flags missing properties with copy-paste fixes. If you're not, the validators above are free and worth running on your top 10 pages once a quarter.
#The pattern
Schema markup is not about ticking a box. It's about giving the machines reading your site a clean, structured statement of what's on the page. The five types above account for almost all of the practical lift. Everything else is optimization theatre unless you have a specific reason.
Related field notes
January 18, 2026 · 3 min
dateModified. The most underrated SEO lever in 2026.
Most SaaS sites set dateModified once at publish time and never touch it again. The cost is real, the fix is cheap, and the trap of "fake date refresh" is worth understanding before you reach for it.
November 20, 2025 · 3 min
FAQ pages still work in 2026 (if you build them right)
Google deprecated FAQ rich results in 2023. Many teams concluded FAQ markup was dead. The opposite turned out to be true for AI engines, but only if your FAQ is actually a FAQ. Here's the distinction that matters.
Share or discuss
New posts, no spam. Roughly monthly. Unsubscribe with one click.