Schema Markup Types Explained for AI-Built Sites

Why "800+ types" is the wrong number to worry about

Schema.org's full vocabulary covers an enormous range of subject matter, from cars and recipes to music albums and medical conditions, because it's built to describe the entire open web, not any one site. Almost none of that range applies to a typical small business or content site. The practical question isn't "which of the 800 types should I learn," it's "which handful actually describes the pages I'm building," and that list is short and stable across nearly every AI-built site this guide covers.

The six types that cover almost every small site

Organization describes your business itself: name, logo, contact details, social profiles. LocalBusiness is a more specific version of it, meant for a business with a physical location customers actually visit, operating hours and all. Article or BlogPosting marks up a genuine written piece, a guide, a post, an explainer. FAQPage marks up a real, visible question-and-answer section, not a summary of the page's content written as fake questions. HowTo marks up genuinely sequential, step-by-step content, not just a page that mentions steps somewhere in passing. BreadcrumbList marks up your site's navigation trail, and WebApplication, the more specific type Schema.org defines for a browser-based tool (a subtype of the broader SoftwareApplication), applies specifically to an interactive tool a visitor can actually use on the page, not a page that just describes one.

Can you actually add custom schema on your AI builder?

Lovable and Bolt.new both give you real code access to your project, so adding a <script type="application/ld+json"> block with your own JSON-LD is the same as adding any other custom code, whether that's directly in a page's HTML or in the component that renders it. Base44 also supports this, confirmed directly against its own documentation: the Code tab lets you edit the generated index.html file directly, which is where a custom JSON-LD block goes, and a separate toggle in the SEO & GEO dashboard's Advanced Settings turns off Base44's own automatic structured-data injection so your hand-added markup doesn't get overwritten by the platform's default. Where the block goes, and four checks that prove it's really there, are covered in How to Add Schema Markup to an AI-Built Site.

The one mistake that undoes all of this

Schema markup that doesn't match what's actually visible on the page. Google's own structured data guidelines are explicit that markup has to reflect real, visible content, not a summary written specifically for the schema block. A FAQPage listing a question nowhere on the visible page, or a HowTo describing steps the page's own text doesn't actually walk through, is the single most common way a page's structured data stops helping and starts working against it. The safest practice is treating the visible text as the source of truth and writing the schema to match it exactly, not the other way around.

Checking it actually works

Google's Rich Results Test checks a live URL or a pasted code snippet directly and reports exactly what structured data it found on the page, along with any errors or warnings. Worth running after adding or changing schema on a page, not only once at launch, since a later content edit (a reworded FAQ answer, a renamed section) can silently break the match between the visible page and its markup without anyone noticing until a test catches it.