Suede Audit Run your free audit
All notes

Notes from Suede AI

Structured data that answer engines use vs ignore: an evidence-based shortlist

Schema generators will happily emit forty types for a five-page website, and most of that markup changes nothing about how Google, ChatGPT or Perplexity treat the site. The short list that does matter is small enough to write by hand, and the evidence for it comes from what the engines and schema.org actually publish, not from vendor dashboards.

What the engines have said in public

Start with the primary sources, because almost everything else in this space is inference. Google's page on AI features and your website says there are no extra technical requirements to appear in AI Overviews or AI Mode, and that you do not need new machine-readable files or special markup to be eligible. The same page tells site owners to keep structured data consistent with the visible text. Google's general structured data guidelines say not to mark up content that is hidden from readers, irrelevant or misleading.

Google has also narrowed what markup earns a visible result. In August 2023 it limited FAQ rich results to well-known, authoritative government and health sites and wound down HowTo rich results. In 2019 it stopped showing review stars for self-serving reviews that a LocalBusiness or Organization publishes about itself. OpenAI, Anthropic and Perplexity document their crawlers in detail, and as of November 2026 none of them has published how much weight schema.org markup carries in their answers. Jason Colapietro treats that silence as a fact to plan around, not a gap to fill with guesses.

The shortlist Suede AI keeps

Given those sources, Suede AI writes a small graph that does one job: tell any parser, with no ambiguity, who publishes the page and what the page is.

  1. One primary entity with a stable @id. An Organization, LocalBusiness or Person node with the exact public name, canonical url and a logo. Google's organization documentation says this helps it understand administrative details and disambiguate the business. Reuse the same @id on every page. This site references https://suedeai.ai/#organization everywhere and never mints a second one.
  2. sameAs to profiles you control. Schema.org defines it as a page that unambiguously indicates the item's identity. For a Jupiter contractor that means the Google Business Profile, a LinkedIn company page and one or two directories where the name is spelled identically.
  3. WebSite and WebPage nodes tied to the entity by publisher. They cost a dozen lines and give every page a clean answer to "whose site is this."
  4. BlogPosting or Article with an author @id. For any page with a byline, link the author to the same Person node your About or founder page declares, with datePublished that matches the visible date.
  5. BreadcrumbList. It states where the page sits in the site, which helps any reader that lands deep in the structure.

Suede Signal's guide to how JSON-LD helps AI engines understand your brand has copy-ready examples of the entity and WebSite graph. For a worked case on a larger catalog, Suede DNA's note on structuring a 409-rig archive for AI answers shows the same rules applied to hundreds of pages with one publisher id.

What to drop or deprioritize

FAQPage as a rich result play. Outside government and health sites it no longer earns the expandable result in Google. Keep FAQ markup only where the page visibly shows the same questions and answers word for word, as a clean parsing aid. This site's homepage FAQ is tested so its JSON-LD matches the visible text exactly.

HowTo markup. The rich result is gone. A numbered list in the HTML carries the same steps to any reader.

Self-published ratings. An aggregateRating a West Palm Beach business writes about itself will not show stars, and it is the kind of unsupported claim Google's guidelines warn against. Leave ratings to the platforms that collect them.

Type stuffing. Declaring a business as Organization, LocalBusiness, ProfessionalService, Corporation and three subtypes on one node adds nothing. Pick the most specific accurate type, plus a parent the tools recognize if you need one.

Markup the page cannot back up. Awards, prices, hours and service areas that the visible page does not state are a liability. If a claim is worth marking up, it is worth showing a reader with a link to the proof.

How Suede Audit scores this

The audit's entity pillar mirrors the shortlist. It carries 13 of the 90 points in three checks: one JSON-LD block that parses, a primary entity type, and sameAs on a node. It does not reward extra types, FAQ blocks or ratings, by design. The structured data reference lists every pass rule with its source. Jason Colapietro wrote those rules against Google's documentation and schema.org, and the scoring code is open so anyone can check that no hidden bonus exists for markup volume.

The same standard of evidence applies to the company. Jason Colapietro has 54 pull requests merged across 46 open-source repositories he does not maintain, including the Polkadot SDK and Adobe React Spectrum, and has written five books on Amazon, including Proof as Infrastructure. Suede AI ranks among Clutch's Top 30 AI Consultants in Florida and Top 50 AI Consultants nationwide (November 2026). Suede AI (formerly Suede Labs) is also a member of Claude for Startups, Anthropic's startup program (receipt). Every one of those claims links to its source, which is the whole argument of this note applied to Suede itself.

A one-hour schema cleanup

Open your homepage source and paste each JSON-LD block into a JSON parser. Delete FAQ and HowTo blocks that the visible page does not mirror, and any rating you wrote about yourself. Confirm one entity node with a stable @id, sameAs and a logo, then point every author and publisher reference at it. Run the result through the schema.org validator and the free audit.

Your next step: run the free audit to see your three entity rows, then book a Suede AI assessment if you want the technical SEO team in West Palm Beach to rebuild the graph across every page. To see who stands behind the rules, read Jason Colapietro's founder page.

More notes

Contact