CiteFlow
Sign Up
← / Blog

Schema

schemajson-ldai-searchtechnical-seo

Schema.org Markup for AI Search: Complete 2026 Guide

Which schema types AI engines actually weight, JSON-LD patterns that get cited, and platform-specific tweaks for ChatGPT, Perplexity, and Google AI Overviews.

5 min read

TL;DR

  • AI search engines treat JSON-LD as text on the page. The schema you ship is read directly into the model context, not just used as a downstream ranking hint.
  • A study cited by hashmeta.ai puts properly schema-marked pages at ~2.5x higher likelihood of appearing in AI-generated answers; an analysis of 73 sites showed 3.2x more AI citations on schema-instrumented pages.
  • Five schema types do almost all the work: Article / BlogPosting, FAQPage, HowTo, Organization, and Product.
  • Field priorities: sameAs, mainEntity, speakable, author with linked profile, datePublished and dateModified.
  • Validate with Google's Rich Results Test and the schema.org validator. Then read the rendered JSON-LD yourself — if it does not read like a clean fact sheet, neither AI engine will treat it as one.

Which schema types AI engines weight most

The honest list, in rough order of payoff per hour of work:

  1. Article / BlogPosting — the canonical signal for editorial content. Sets author, dates, headline, image. The freshness layer in both Google's AI Overviews and Perplexity reads dateModified literally.
  2. FAQPage — when you have a real Q&A block on the page, mark it up. ChatGPT and Perplexity both treat the resulting JSON-LD as readable text, so the dual signal of visible HTML plus structured Q&A compounds.
  3. HowTo — step-by-step instructional content. Maps cleanly to the way LLMs synthesize procedural answers.
  4. Organization — once per site, on your About or homepage. Pins your name, url, sameAs profiles, and contact data into the entity graph the engines build of you.
  5. Product — for SaaS and ecommerce. Includes aggregateRating, offers, and brand. Drives both rich results and AI Overview product references.

Other types (Event, Recipe, Review, JobPosting) matter in their verticals but do not generalize. Start with the five above.

JSON-LD patterns that work

Article (BlogPosting variant)

{
  "@context": "https://schema.org",
  "@type": "BlogPosting",
  "headline": "How to Rank in Google AI Overviews: 2026 Playbook",
  "description": "Tactical guide to ranking in Google AI Overviews — page templates, schema, freshness signals, and how to track citation share.",
  "datePublished": "2026-05-08T00:00:00Z",
  "dateModified": "2026-05-08T00:00:00Z",
  "author": {
    "@type": "Person",
    "name": "Jane Doe",
    "url": "https://example.com/about",
    "sameAs": ["https://www.linkedin.com/in/janedoe"]
  },
  "publisher": {
    "@type": "Organization",
    "name": "Acme",
    "url": "https://example.com",
    "logo": {
      "@type": "ImageObject",
      "url": "https://example.com/logo.png"
    }
  },
  "image": "https://example.com/blog/ai-overviews-cover.jpg",
  "mainEntityOfPage": {
    "@type": "WebPage",
    "@id": "https://example.com/blog/ai-overviews-2026"
  }
}

FAQPage

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [
    {
      "@type": "Question",
      "name": "Do I need to publish 6,000-word ultimate guides to get cited?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "No. The empirical pattern favors several interlinked 1,200-2,000-word articles over one mega-page."
      }
    },
    {
      "@type": "Question",
      "name": "Does blocking GPTBot affect AI Overviews?",
      "acceptedAnswer": {
        "@type": "Answer",
        "text": "GPTBot is OpenAI's crawler. AI Overviews are powered by Googlebot and Google-Extended. The latter is the lever you actually pull."
      }
    }
  ]
}

Two patterns that show up in cited pages but rarely in guides: each acceptedAnswer.text is one or two complete sentences that stand alone if quoted, and the Question.name mirrors how a human would actually phrase the query.

Field-level priorities

A few fields punch above their weight:

  • sameAs — on both Person and Organization. Links your entity to LinkedIn, Wikipedia (where applicable), Crunchbase, GitHub, and authoritative profiles. Drives entity disambiguation across engines.
  • mainEntity — used on FAQPage, QAPage, HowTo. Tells the parser what the page is fundamentally about. Skip it and the page becomes a generic blob.
  • speakable — flags the passages most suitable to be read aloud (or quoted). Voice assistants use it; AI Overviews appear to weight it lightly. Inexpensive to add.
  • author — always a Person, ideally with a url and sameAs. Anonymous content is downweighted in E-E-A-T-driven retrievers.
  • datePublished and dateModified — both in ISO 8601, both UTC. The freshness layer parses these directly.

What to leave off: keywords (largely ignored), inLanguage unless you serve multilingual variants, custom extensions outside schema.org's vocabulary (parsers silently drop them).

Validating

Two tools you should run on every important page:

  1. Google Rich Results Test — catches structural errors and reports which rich-result types are eligible.
  2. Schema.org validator — stricter on vocabulary conformance. Catches the cases where Google accepts the markup but it is technically wrong.

Then do the third thing nobody does: open the rendered JSON-LD in your browser dev tools and read it. If it does not narrate the page as cleanly as a sentence would, fix the data, not the schema.

Platform-specific tweaks

  • ChatGPT. Pulls heaviest from Article and FAQPage. Plain-language values matter more than nested entity graphs. Avoid over-engineering.
  • Perplexity. Rewards entity-rich markup — Organization with thorough sameAs, Person with linked profiles, named comparators in Product markup. Helps the entity ranker confirm you exist as a known entity.
  • Google AI Overviews. Rewards the full E-E-A-T stack — Article with author, publisher, dates, plus mainEntityOfPage. Freshness fields matter most here.
  • Gemini. Largely overlaps with Google AI Overviews since it shares the underlying index. Add speakable for voice-driven Gemini surfaces.

FAQ

Will adding schema alone get me cited?

No. Schema is a force multiplier on otherwise good content. A schema-marked page that does not answer the query still does not get cited; a schema-marked page that does answer it cleanly gets cited noticeably more often than the same content without schema.

JSON-LD, Microdata, or RDFa?

JSON-LD. The other two are technically valid but no longer recommended by Google, and AI parsers handle JSON-LD more reliably. Migrate any legacy Microdata to JSON-LD.

Should I mark up every page?

Pillar pages, comparison pages, how-to pages, and FAQ pages — yes. Thin index pages, tag archives, and login pages — no. Apply the rule that schema goes on pages whose content you want lifted into an AI answer.

Sources

Want a schema audit on your most important pages? CiteFlow checks JSON-LD coverage, validates against schema.org, and flags the AI-relevant fields you are missing.

Related reading