The 14-Point Schema Markup Checklist for Google AI Overviews (2026)

Pages with valid schema markup are 2–4x more likely to appear in Google's AI Overviews, and a 2026 study from Discoverability found websites with comprehensive structured data see up to a 44% lift in AI search citations. Schema is the cheapest GEO lever you have — one afternoon of dev work, no content rewriting, compounding returns. This post is the 14 schema types we ship across client sites in 2026, in priority order, with copy-paste JSON-LD examples. Read top-down: ship the first 5 today, the next 5 this month, the last 4 when you are ready. ## TL;DR — what to ship in what order Ship FAQPage, Article, BreadcrumbList, Organization, and Person schema first — those five cover 80% of the AI Overview impact. Add HowTo, Product, Service, Review, and AggregateRating next — these matter for transactional and instructional queries. Finally add Speakable, VideoObject, Course, and SoftwareApplication — these are vertical-specific and open up voice and NotebookLM-style channels. Validate every change at Google's Rich Results Test and Schema.org's validator. The 2026 AI engines (Google AI Overviews, Perplexity, ChatGPT, Claude) all parse JSON-LD reliably — XML-style microdata is dying and we no longer ship it. ## Why schema matters even after FAQ rich snippets died Google deprecated FAQ rich snippets in May 2026 — they no longer appear in classical SERPs. The instinct of many SEO teams was to rip FAQPage schema out of their pages. That is backwards. ChatGPT, Perplexity, Claude, and Google AI Overviews all still parse FAQPage schema to identify, extract, and cite content. The rich snippet was the visible reward; the citation pull is the invisible one, and the invisible one is now bigger. ## The 14 schema types, in priority order ### 1. Article / BlogPosting — the foundation Every blog post gets this. It tells AI engines what is editorial content, who the author is, when it was published, when it was last updated, and what entity (your Organization) published it. Without Article schema, your blog is just a soup of text to a crawler. ### 2. FAQPage — the AI citation magnet Every page with 3+ question-format H3s gets FAQPage schema. Use real customer questions — generic "What is X?" templates get filtered. ### 3. BreadcrumbList — your information architecture Tells AI engines how your site is organized. Adds the URL trail to AI Overview attribution chips. ### 4. Organization — your entity card One sitewide Organization block on the homepage and the About page. The most important fields: name, url, logo, sameAs (your social profiles), founder, address. We always include LinkedIn, X, YouTube, and a founder Person object linking out to our founder's personal site — the more entity signals, the better Google's knowledge graph understands you. ### 5. Person — for every author Every blog author gets a Person schema. AI engines use the Author entity to attribute content and weigh credibility. Without it, "Softechinfra Team" is anonymous and rankings suffer. ### 6. HowTo — for every step-by-step guide Tutorial posts and walkthroughs get HowTo schema. Each step becomes a discrete citation candidate. Especially powerful in AI Overviews for "how do I..." queries. ### 7. Product — for every service or product page Even for a services firm, treating each service as a Product enables ratings, pricing, and availability fields to surface in AI Overviews. ### 8. Service Used in parallel with Product for services. Defines provider, serviceType, areaServed, and links to relevant case studies. Google treats Service schema as a strong entity signal for local + AI search. ### 9. Review and AggregateRating If you have 5+ legitimate client reviews, AggregateRating with ratingValue and reviewCount becomes one of the strongest AI Overview signals for service buyers. Fake or scraped reviews trigger penalties — only use this if reviews are real and verifiable. ### 10. Speakable — voice and NotebookLM Speakable schema marks specific page sections as suitable for text-to-speech extraction. Initially designed for Google Assistant, it became relevant again in 2025 when NotebookLM-style audio overviews emerged as a distribution channel. You define a CSS class on the elements you want read aloud (typically the TL;DR paragraph and the answer paragraphs under each H2), then reference those classes in the Speakable schema. Wrap your TL;DR in <div class="speakable-tldr"> and you are done. ### 11. VideoObject — for embedded video Every embedded YouTube or self-hosted video on a content page gets VideoObject schema. It tells AI engines what the video is about and where in the timeline the key moments live. Especially useful for AI Overview video carousels. ### 12. Course Edtech and training pages get Course schema. Our in-house product PenLeap uses Course schema heavily for its exam-prep modules — each module becomes a discoverable entity. For Softechinfra services this is mostly relevant on workshop and training landing pages. ### 13. SoftwareApplication — for tool reviews and SaaS If you write about a software tool, wrap each named tool in SoftwareApplication schema with applicationCategory, operatingSystem, and offers. AI engines pull these into comparison answers verbatim. ### 14. ImageObject and the visual entity layer Every important image on a page (the hero, charts, screenshots) gets ImageObject schema with contentUrl, creator, creditText, license. This is part of Google's Image Discovery infrastructure but increasingly fed into AI Overviews for visual answers. ## The DIY ship plan (3 days) Pre-flight checklist before you start the dev push:

• You have admin access to your CMS or _document file (Next.js, Astro, WordPress)
• You can validate at Google Rich Results Test on staging URLs
• You have author bio data (real photo, LinkedIn URL, X handle, role) for every Person schema
• You have 5 to 7 real customer questions per priority page (pulled from sales call notes)
• You have your Organization sameAs list ready (LinkedIn, X, YouTube, founder personal site)

## Common schema mistakes that get you penalized Hidden content in schema. Google's structured data policy requires the content in your schema to be visible on the page. We see clients hide 30 fake FAQ questions in JSON-LD with only 3 visible on the page. This triggers a manual action. Markup that contradicts visible content. If your AggregateRating says 4.9 stars and your visible reviews show 3.2, that is a penalty waiting. Schema on doorway pages or thin content. Schema cannot save a 200-word stub. Google's algorithmic filters explicitly demote pages with rich schema and thin content. Multiple Organization schemas on one page. Pick one. Multiple conflicting Organization blocks confuse the entity graph and we have seen citations drop measurably when this is fixed. Stale dateModified. If you say dateModified: 2026-04-03 but you have not touched the page, AI engines down-weight the freshness signal once they realize you are gaming it. Only update dateModified when you actually update content. ## A real example A 14-person Coimbatore textile-tech SaaS client came to us with 38 blog posts and zero schema. Indexed in Google, never cited in AI Overviews. We shipped the 5 priority schemas across the whole site in a single 6-hour push, validated everything, and pinged the Google Indexing API. In 4 weeks, they appeared as a cited source in 7 AI Overview answers for their target queries. No content changed. Schema alone, 4 weeks. This is the cheapest GEO lever we know. ## FAQ ### Does schema help if my content is bad? No. Schema amplifies the signal of good content; it does not invent signal. A thin or unhelpful page with full schema will not be cited. Fix the content first, schema second. ### Should I use JSON-LD or Microdata? JSON-LD only. All major AI engines (Google, Bing/ChatGPT, Perplexity, Claude) parse JSON-LD reliably. Microdata and RDFa are legacy formats — we have removed them from every client site we work on without any measurable loss. ### Where do I put the JSON-LD block on the page? <head> is best for AI extractors and matches Google's recommendation. Some teams put it before </body> for build-system reasons; it works but <head> is cleaner. Never put it inside <noscript>. ### How often does Google re-parse my schema? Every Googlebot crawl. For a fresh post, that is hours. For an old page, it can be weeks. Use Google Search Console's URL Inspection to force re-crawl on a critical page. ### What is the single highest-impact schema to ship today? FAQPage with 5–7 real customer questions on your top 10 pages. Half a day of work, immediate AI citation lift, near-zero downside. ### Do I need schema if I use a CMS that ships some by default? Probably yes. WordPress + Yoast ships limited Article schema and basic Organization. Most CMSes miss FAQPage, HowTo, Speakable, and Service entirely. Audit what your CMS emits with Google's Rich Results Test, then layer in the gaps with a custom snippet. ### Does Schema.org's spec change frequently? Yes, twice a year typically. Most changes are additive (new types). The 14 in this list are stable as of May 2026 and unlikely to change before 2027. Subscribe to the Schema.org release notes if you ship schema professionally.