Schema Markup for AI Search: Which Types Actually Help

Schema markup is JSON-LD code that tells AI engines what your content is and which parts answer what, and the types that move the needle for AI citation are Article, FAQPage, HowTo, Organization, and Person. Google says no schema is "required" for AI Overviews, but the evidence points to a real edge. One 2026 ranking-factors study found pages with Article, FAQPage, and HowTo schema saw a +73% selection rate for AI Overview citation, and author (Person) schema correlated with roughly a 3x lift in AI answer appearances (BrightEdge).

The reason is mechanical. AI engines retrieve and reassemble passages, and structured data hands them clean, labeled chunks: this is the author, this is the publish date, this is a question and its answer, this is step three. You are removing ambiguity, which makes you safer to quote. Think of schema as subtitles for a machine reader.

Why schema matters more for AI than for classic SEO

In traditional search, schema mostly earned you rich results, those star ratings and FAQ dropdowns in the SERP. Google has since scaled back some of those displays, which led some people to declare schema dead. It is not. For AI search, schema does something different and more durable: it helps the model parse entities and extract answers reliably. The display benefit shrank; the parsing benefit grew. For a GEO strategy, that is the part you care about.

The types worth implementing, ranked

1. Article (or BlogPosting). The baseline for every post. Include author (linked to a Person), datePublished, dateModified, and publisher. This is what carries your freshness and authorship signals, two of the heaviest-weighted factors in AI selection.
2. Person (author). A real, named author with a bio and a sameAs link to a verifiable profile like LinkedIn. This is the cheapest high-impact win, and most sites still ship anonymous posts that leave the author signal on the floor.
3. FAQPage. For genuine question-and-answer pairs. Each Q&A becomes a direct citation candidate. Do not fake it on unrelated content.
4. HowTo. For step-by-step guides. Each step becomes an independently extractable passage, which is ideal for procedural queries like "how to set up X."
5. Organization. Site-wide, so your brand reads as a clear entity with a consistent name, logo, and sameAs profiles. This anchors entity recognition across engines.
6. Product or SoftwareApplication. For your tool and pricing pages, so AI can extract specifics like price, features, and ratings when someone asks about your category.

Copy-paste starting points

Article schema, the one to add first:

{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "How to Get Cited by ChatGPT",
  "datePublished": "2026-06-24",
  "dateModified": "2026-06-24",
  "author": {
    "@type": "Person",
    "name": "Your Name",
    "url": "https://www.linkedin.com/in/yourprofile",
    "sameAs": ["https://www.linkedin.com/in/yourprofile"]
  },
  "publisher": {
    "@type": "Organization",
    "name": "Okara",
    "logo": { "@type": "ImageObject", "url": "https://okara.ai/logo.png" }
  }
}

FAQPage schema, for pages that genuinely answer questions:

{
  "@context": "https://schema.org",
  "@type": "FAQPage",
  "mainEntity": [{
    "@type": "Question",
    "name": "Does schema help with AI Overviews?",
    "acceptedAnswer": {
      "@type": "Answer",
      "text": "Google says no schema is required, but pages with Article, FAQPage, and HowTo schema show measurably higher AI Overview selection rates."
    }
  }]
}

HowTo schema, for step-by-step guides, follows the same pattern with a step array, where each HowToStep has a name and text. Each step then becomes its own extractable passage.

Rules that keep schema from backfiring

• The markup must match the visible content. Schema describing content that is not on the page is a quality violation and can get you penalized. The JSON-LD should mirror what a reader sees.
• Use JSON-LD, in the page head or body. It is the format Google recommends and the easiest for engines to parse. Skip microdata and RDFa.
• Validate it. Run pages through Google's Rich Results Test and Schema.org's validator before shipping. A typo in your JSON-LD silently breaks the whole block.
• Do not slap FAQPage onto everything. It works when the page genuinely answers questions. Forced FAQs read badly to humans, and engines have learned to discount markup that does not match real Q&A content.
• Keep dateModified honest. Updating the date without updating the content is the kind of thing that erodes trust if a page is obviously stale. Refresh the content, then bump the date.

A note on what schema does not do

Schema is an amplifier, not a substitute. It helps engines parse a good page; it cannot rescue a thin one. If your answer is buried, vague, or unsourced, perfect markup will not get you cited. Get the content structure and sourcing right first, then add schema to make a strong page machine-clean. The order matters: structure, then sourcing, then schema.

Where Okara fits

Schema is exactly the kind of fix that is simple to describe and tedious to implement consistently across a whole site, which is why most sites do it partially or not at all. Okara's SEO agent flags missing or broken schema as part of its continuous audit, and its coding agent can implement technical fixes like JSON-LD without you editing templates by hand. So the +73% selection-rate opportunity actually gets shipped, page by page, instead of staying on a to-do list. Point it at your site to see which pages are missing the schema that matters.

Frequently asked questions

Is schema required to appear in AI Overviews? No. Google states it is not required. But independent studies consistently find a selection-rate edge for pages that use it, so treat it as a strong advantage rather than an optional extra.

Which single schema type should I add first? Article with a linked Person author. It covers authorship and freshness in one move, and author signals correlate with a large lift in AI answer appearances.

Will FAQ schema still show rich results in Google? Google has scaled back FAQ rich results in regular search, but the schema still helps AI engines extract clean Q&A pairs, which is the goal for GEO.

Does schema help with ChatGPT and Perplexity, or just Google? Both. ChatGPT's and Perplexity's crawler guidance both point to structured data as a way to help them understand and match content.

Can bad schema hurt me? Yes. Markup that does not match the visible page is a quality violation, and invalid JSON-LD simply gets ignored. Validate before publishing.