Schema markup is no longer a nice-to-have. In 2025 citation studies, structured data accounts for roughly 10% of how Perplexity ranks sources, and pages with three or more FAQ question nodes get cited nearly twice as often. So we built a fixed 14-point JSON-LD checklist and now ship it on every client site we touch. This post is that checklist — every item, why it matters, and a copy-paste example you can validate today.
The Answer in 60 Words
The 14 schema types we ship: Organization, WebSite with SearchAction, BreadcrumbList, BlogPosting or Article, FAQPage, Person for authors, Service, Product with Offer, Review and AggregateRating, HowTo, VideoObject, ImageObject, LocalBusiness where relevant, and a sameAs entity-graph linking your social profiles. Each is valid JSON-LD, validated in Google's Rich Results Test, and placed in the page head. FAQPage and Organization carry the most GEO weight.
Why This Matters Now (August 2025)
AI search engines read structured data to understand what your page is and whether to trust it. Independent 2025 work found schema contributing around 10% of Perplexity's ranking factors, and pages with valid markup appearing 2 to 4x more often in Google's AI Overviews. We turned the ad-hoc "add some schema" advice into a repeatable 14-point checklist so nothing gets missed. We covered the foundational three-file setup in our llms.txt, schema, and sitemap guide.
The 14-Point Checklist
Here is the full list with the why for each:
A Copy-Paste FAQPage Block
This is the single most valuable item, so here is a runnable example. Place it in the page head, swap in your real questions, and validate:
Build the JSON-LD with a top-level type of FAQPage and a mainEntity array, where each entry is a Question with a name and an acceptedAnswer Answer carrying the text. Keep each answer 40 to 60 words, and make sure it matches the visible on-page answer word for word — mismatches between schema and visible text are a known quality flag.
The 4 Schema Mistakes We Fix Most Often
Most sites we audit do have some schema. The problem is rarely absence — it is errors that quietly make the markup worthless. These four account for the bulk of what we repair.
1. Schema that does not match the visible page. A FAQPage block with questions that do not appear in the on-page text. AI engines and Google both treat the mismatch as a quality flag. The fix: generate the schema from the visible FAQ, not separately, so they can never drift apart.
2. Multiple conflicting Organization blocks. A theme injects one, a plugin injects another, and a developer hand-codes a third — three different logos and names for the same company. Engines cannot tell which is canonical. Consolidate to one authoritative Organization block, sitewide.
3. Invalid JSON that silently fails. A trailing comma, an unescaped quote, a missing bracket — and the whole block is ignored. It looks present in the source but contributes nothing. The fix: validate every block before deploy, every time. Never assume it parses.
4. Stale dateModified that never changes. A page stamped "updated" two years ago, or worse, auto-stamped to today on every build without any real edit. Both undercut the freshness signal. Update the timestamp only when the content genuinely changes, and do change content on a real cadence.
Validation: The Three-Tool Pass
| Tool | What it checks | When to use |
|---|---|---|
| Google Rich Results Test | Eligibility for rich results, parse errors | Before every deploy |
| Schema.org Validator | Strict schema validity | When debugging warnings |
| Search Console | How Google actually reads it in production | Weekly, post-deploy |
The Deploy Checklist
- Organization + WebSite + sameAs shipped sitewide, once
- BreadcrumbList on every deep page
- BlogPosting/Article on every content page with honest dateModified
- FAQPage with 5–7 real questions, matching the visible FAQ block exactly
- Person schema linking each author to a real bio
- Service and Product+Offer on commercial pages, prices in INR
- Review/AggregateRating only where real reviews exist
- Every block validated in Google's Rich Results Test before deploy
The Order We Ship Them In (Highest Impact First)
You do not ship all 14 at once on day one. We sequence by impact-to-effort so the GEO gains start landing in week one, not week six. Here is the order we follow on a typical client site.
This sequence means a client sees their Organization entity and FAQ answers recognised within the first crawl cycle, while the long tail of supporting schema fills in behind. Front-load the two highest-impact types and you do not wait a month to see movement.
What 14-Point Schema Did for a Jaipur Client
A Jaipur interior-design studio came to us indexed but invisible in AI search. Their site had zero structured data. We shipped the full 14 points: Organization with founder and sameAs, LocalBusiness with their studio address, Service schema on each offering, BlogPosting plus FAQPage on every project write-up, and ImageObject on their portfolio shots. Within nine weeks they appeared in Google AI Overviews for "interior designers in Jaipur for small homes" and got their first Perplexity citation. The schema was not the only change — content mattered too — but it was the missing piece that made their good content legible to machines. The same discipline runs through our SEO services and the engineering on ChipMaker Hub.
When You Do Not Need All 14
A pure blog with no products, no physical location, and no video does not need Product, LocalBusiness, or VideoObject — forcing them in adds noise and validation errors. Ship the relevant subset: Organization, WebSite, BreadcrumbList, BlogPosting, FAQPage, and Person cover most content-only sites. Match the schema to what the page actually is; irrelevant markup is worse than none. We later expanded this thinking in our AI Overviews schema deep-dive.
FAQ
Which schema type matters most for AI citations?
FAQPage and Organization. FAQPage turns each Q&A into a machine-readable answer candidate — pages with three or more question nodes get cited roughly twice as often. Organization anchors your entire site as a recognisable entity that AI engines can trust and attribute.
Does FAQPage schema still work after Google deprecated the FAQ rich snippet?
Yes. The Google rich result is separate from the underlying value. Even without the snippet, FAQ schema remains a top GEO signal because ChatGPT and Perplexity read it directly to find quotable answers. We keep it on every content page.
Where should JSON-LD go — head or body?
Either works for parsing, but we standardise on the page head for sitewide blocks (Organization, WebSite) and allow per-page blocks (BlogPosting, FAQPage) in head or just before the closing body tag. Consistency matters more than location; validate after placing.
How do I validate schema markup?
Run every block through Google's Rich Results Test before deploy, use the Schema.org validator for strict checking when debugging, and watch Search Console weekly to see how Google reads it in production. Fix any error or warning before it reaches users.
Can bad schema hurt my rankings?
Yes, in two ways. Invalid JSON-LD can be ignored entirely, wasting the effort. Fabricated Review or AggregateRating markup can trigger a Google manual action that damages your whole domain. Only mark up what is true, and validate everything.
How long until schema affects AI visibility?
Typically 3 to 6 weeks once the pages are recrawled. Schema is necessary but not sufficient — it makes good content legible to machines. Pair it with extractable answers and honest freshness, and the combined effect shows up faster than schema alone.
Want the Full 14-Point Schema Shipped on Your Site?
We audit your current markup, ship all 14 schema types where relevant, validate every block in Google's Rich Results Test, and hand you a maintenance checklist. Fixed scope, 5 working days for a standard site. Email contact@softechinfra.com or book a call.
Book a Schema Audit
