Pages with valid schema markup are 2-4x more likely to appear in Google AI Overviews, and a measured study by Schema App found that sites implementing Organization plus FAQPage schema showed +37% citation rate in AI Overviews vs sites with no schema. We deployed a 14-schema bundle across four softechinfra.com pages on October 14, 2025. By November 13, those four pages had moved from 0 ChatGPT citations to 11 across our 18-query baseline. This post is the exact JSON-LD bundle, validated and copy-paste ready, with the 30-day metrics by page.
TL;DR — the 14-schema stack in one paragraph
The bundle we ship on every page is: Organization, Service, ProfessionalService (deprecated but still parsed), LocalBusiness, FAQPage, HowTo, Article (or BlogPosting), Person (author), BreadcrumbList, WebPage, WebSite, SiteNavigationElement, SpeakableSpecification (beta but Google-supported), and ImageObject. Tier 1 (Organization, FAQPage, HowTo, Article) does most of the work. Tier 2 (the rest) makes your entity graph clean for LLMs that build site-level context. Total file size adds about 4-8KB per page — well inside any performance budget.
Why this matters now (Q4 2025 trigger)
In May 2025 Google quietly stopped showing FAQ rich results in regular SERPs for most queries — the schema is now used primarily as a signal for AI Overviews and AI Mode rather than for visible blue-link snippets. That changed the cost-benefit calculation. You no longer add FAQPage schema for the visible click-through-rate lift; you add it because AI engines treat it as the canonical "this site has answers to these questions" signal. The same shift happened for HowTo. Speakable, still in beta, is now actively used by voice-mode AI assistants. Schema is no longer "nice to have for snippets" — it is the primary structured-data signal AI engines parse for citation candidates.
The full 14-schema bundle (copy-paste, validated)
We split this into Tier 1 (must-have on every page) and Tier 2 (sitewide, deploy once). All schema below has been validated against schema.org validator and Google's Rich Results Test on November 10, 2025.
Tier 1, item 1 — Organization (sitewide)
Place in the <head> of every page or once in your root layout component.
{
"@context": "https://schema.org",
"@type": "Organization",
"@id": "https://www.softechinfra.com/#organization",
"name": "Softechinfra",
"url": "https://www.softechinfra.com",
"logo": "https://www.softechinfra.com/logo.png",
"founder": {
"@type": "Person",
"name": "Vivek Kumar",
"url": "https://viveksinra.com"
},
"sameAs": [
"https://www.linkedin.com/company/softechinfra",
"https://twitter.com/softechinfra",
"https://github.com/softechinfra"
],
"contactPoint": {
"@type": "ContactPoint",
"email": "contact@softechinfra.com",
"contactType": "customer service",
"areaServed": "IN"
}
} The @id anchor lets every other schema on your site reference the Organization without redefining it — critical for entity graph consistency.
Tier 1, item 2 — Service (per service page)
{
"@context": "https://schema.org",
"@type": "Service",
"name": "AI Automation",
"provider": { "@id": "https://www.softechinfra.com/#organization" },
"areaServed": { "@type": "Country", "name": "India" },
"serviceType": "AI workflow automation, n8n development, chatbot integration",
"description": "We design and ship n8n workflows, WhatsApp chatbots and LLM integrations for Indian SMBs. Typical engagement: 7-21 days, fixed scope.",
"offers": {
"@type": "Offer",
"priceCurrency": "INR"
}
}We name the price range. Vague schema gets ignored; concrete numbers get cited.
Tier 1, item 3 — FAQPage (per content page with FAQs)
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "How much does an n8n self-hosted setup cost in India?",
"acceptedAnswer": {
"@type": "Answer",
"text": "It depends on scope; a Hetzner CX22 (2 vCPU, 4GB RAM) covers typical SMB workloads. Our team can also harden the install and ship the first three workflows."
}
}
]
}Each Question + Answer mirrors a real H3 + paragraph from your visible FAQ block. AI engines cross-check the schema against rendered HTML — fake or hidden questions get penalised.
Tier 1, item 4 — Article / BlogPosting + Person (author)
{
"@context": "https://schema.org",
"@type": "BlogPosting",
"headline": "Article title here",
"datePublished": "2025-11-13",
"dateModified": "2025-11-13",
"author": {
"@type": "Person",
"name": "Vivek Kumar",
"url": "https://www.softechinfra.com/team/vivek-kumar",
"sameAs": ["https://www.linkedin.com/in/viveksinra"]
},
"publisher": { "@id": "https://www.softechinfra.com/#organization" },
"mainEntityOfPage": "https://www.softechinfra.com/blog/your-slug"
} dateModified is non-cosmetic. Update it any time you make a substantive edit. AI engines weight freshness; ChatGPT and Perplexity both show recency bias of 12-18 months on cited sources.
Tier 2 — Speakable, BreadcrumbList, WebSite, SiteNavigationElement, ImageObject
These are the polish layer. Each adds entity-graph clarity for LLMs building a site-level model. Total deploy time: half a day across the site.
{
"@context": "https://schema.org",
"@type": "WebPage",
"speakable": {
"@type": "SpeakableSpecification",
"cssSelector": [".tldr", ".faq-question", ".faq-answer"]
}
} Speakable is still in beta (Google-supported). Mark your TL;DR block, FAQ questions and answers — those are the chunks voice assistants pull. We use a .tldr CSS class on the answer-box paragraph after every H2.
{
"@context": "https://schema.org",
"@type": "BreadcrumbList",
"itemListElement": [
{ "@type": "ListItem", "position": 1, "name": "Home", "item": "https://www.softechinfra.com" },
{ "@type": "ListItem", "position": 2, "name": "Blog", "item": "https://www.softechinfra.com/blog" },
{ "@type": "ListItem", "position": 3, "name": "Schema Markup Stack", "item": "https://www.softechinfra.com/blog/schema-markup-stack-chatgpt-citations-30-days" }
]
}The DIY walkthrough — ship the bundle in one day
This is the four-hour rollout we run for clients on a 5-10 page priority set. Adjust headcount up if you have 50+ pages.
- Organization schema with founder + sameAs deployed sitewide
- Service schema on every /services/[slug] with priceRange in INR
- FAQPage schema mirrors rendered FAQ block on every content page
- Article/BlogPosting + Person (author) on every blog post
- BreadcrumbList + Speakable on long-form pages
- All schema validated at validator.schema.org and Google Rich Results Test
The 30-day metrics from softechinfra.com
Honest numbers, including the page that did not move. We deployed the bundle on October 14, 2025 across four pages and tracked citations across an 18-query baseline through November 13.
| Page | Citations Oct 14 | Citations Nov 13 | Schema added |
|---|---|---|---|
| /services/ai-automation | 0 | 4 (ChatGPT, Perplexity) | Organization, Service, FAQPage, Speakable |
| /services/crm-development | 0 | 3 (Perplexity, AI Overviews) | Organization, Service, FAQPage, HowTo |
| /blog/softechinfra-9-n8n-workflows-2025-and-3-we-killed | 1 | 3 (ChatGPT, Perplexity) | BlogPosting, FAQPage, HowTo, Speakable |
| /services/seo-services | 0 | 1 (Perplexity) | Organization, Service, FAQPage |
| Total | 1 | 11 | — |
The /services/seo-services page underperformed. Our hypothesis: the page is too generic versus narrower competitors who specialise in only-SEO. We are now testing a rewrite that adds a "GEO sub-service" angle and four named case studies — same schema, sharper content. Will report back in the December roundup.
The Hrishikesh test — what to validate before deploy
The four-step pre-deploy QA Hrishikesh runs on every client schema rollout: (1) schema.org validator passes with zero errors and zero warnings; (2) Google Rich Results Test shows the page as eligible for at least one rich result type; (3) every Question text in FAQPage appears verbatim in rendered HTML; (4) curl with the Googlebot User-Agent returns the same JSON-LD as a regular browser (catches client-side rendering bugs that break schema for crawlers).
Common mistakes that kill schema-driven citations
FAQPage schema with hidden questions. Google penalises schema that does not match visible content. If your FAQ is in a closed accordion that requires JS to render, your schema may be ignored. Server-side render the FAQ HTML or use <details> / <summary> so the content is in source.
Multiple Organization schemas with different @id. Use the same @id URL fragment everywhere ("https://www.softechinfra.com/#organization"). Conflicting Organization records confuse the entity graph.
Missing dateModified. Without dateModified, Article schema falls back to datePublished — and AI engines treat the page as stale after 90 days. Update dateModified any time you edit substantively.
Schema in JSON-LD that conflicts with Open Graph. Your og:title and your headline in BlogPosting must match. Mismatch is a quality signal — keep them in sync via a single source of truth in your CMS.
Adding Speakable to the wrong selectors. Speakable should mark the 20-30 second audio chunks: TL;DR, FAQ questions, FAQ answers. Marking your full article body causes voice assistants to read 12 minutes of text — they ignore it instead.
For more on the underlying GEO signals this schema bundle amplifies, see our companion post on the 8 ranking signals AI engines actually use. The schema is the multiplier; those eight signals are the foundation. Our SEO services team ships both together on every engagement.
When schema does not help (and what does)
If your page is genuinely thin (under 800 words, no original numbers, no real examples), schema cannot save it. AI engines use schema as a "canonical answer record" indicator — if the canonical answer is shallow, the page still loses. Rebuild the content first. We turn down clients monthly who want only a "schema deploy" without a content audit. The Princeton three (statistics, citations, quotations) on real content is the prerequisite — schema is the multiplier on top.
A real example — what we shipped for ChipMaker Hub
For ChipMaker Hub (semiconductor design analytics, 30-person team), we deployed the same 14-schema bundle plus a custom DataFeed schema for their published industry datasets. Result: their landing page moved from "not in any AI citation" to being the named source on three Perplexity queries about Indian semiconductor manufacturing capacity within six weeks. Same content. New schema. The schema was not the only change (we also rewrote the answer-first paragraph after each H2) — but the schema was the cheaper of the two changes by an order of magnitude.
FAQ — the schema questions clients keep asking
Do I need every one of the 14 schemas?
No. Tier 1 (Organization, FAQPage, HowTo where applicable, Article/BlogPosting) does about 80% of the lift. Add Tier 2 once Tier 1 is live and stable. Skipping Tier 2 is fine on small sites (under 50 pages).
Will schema work if my site is SPA / client-rendered?
Only if you server-render the JSON-LD or use a pre-render service. Many AI crawlers do not execute JavaScript. Test with curl -A "GPTBot/1.0" and confirm the JSON-LD is in the raw HTML response.
Should I use ProfessionalService or Service?
Schema.org has deprecated ProfessionalService in favour of Service. Use Service. Add the more specific subclasses (LegalService, MedicalBusiness, etc.) only if you fit one cleanly.
How often should I update schema?
Sitewide schema (Organization, WebSite) only when you genuinely change. Per-page schema (Article, FAQPage) any time the underlying content changes meaningfully. Update dateModified on every substantive edit — not on every typo fix.
Does schema help for ChatGPT specifically?
Indirectly. ChatGPT uses Bing's index, and Bing uses schema to understand pages. Bing has confirmed in its Webmaster guide that valid schema improves understanding. The downstream effect on ChatGPT citations is real but harder to attribute cleanly than Perplexity, which is more transparent.
Does FAQPage still get rich snippets in Google?
Mostly no. Google narrowed FAQ rich results in May 2025 to a small set of authoritative health and government sites. The schema is still valuable for AI Overviews and AI Mode — keep deploying it.
Can I use a schema generator instead of hand-writing?
Yes. Merkle's generator and schema.org's own validator are good. Hand-validate the output before deploy — generators sometimes inject deprecated properties.
Want all 14 schemas implemented site-wide?
We deploy the full Tier 1 + Tier 2 bundle on Indian B2B sites in 4-7 working days. Includes audit of current schema, deploy to your stack (Next.js, WordPress, custom), validation against schema.org and Google Rich Results Test, and a 30-day post-deploy citation report.
Get Schema Bundle Quote
