Is Your JSON-LD Helping or Hurting Your AI Visibility?
Reviewed August 2026.
Does your JSON-LD confirm what the page already says, or contradict it?
The Answer Patch Fix Report's Structured Data pillar checks JSON-LD the scoring engine actually validates: LocalBusiness for qualifying local sites, FAQPage quality and match to visible copy, BreadcrumbList on inner pages, absolute URLs, and page-type expectations such as Organization and SoftwareApplication for SaaS. Start with a free homepage score; schema snippets ship in the report when facts allow.
Structured data lets a machine verify entities and facts instead of inferring them from prose. The Fix Report's Structured Data pillar does not score a generic "any JSON-LD is good" checkbox. It validates LocalBusiness when the site is a qualifying local business, flags malformed FAQPage questions, compares schema name and phone to visible copy on local-service sites only, checks BreadcrumbList on non-home pages, requires absolute URLs, and expects page-type markup, LocalBusiness on a local homepage, Organization, WebSite, and SoftwareApplication on SaaS surfaces, Article or BlogPosting on editorial URLs for non-local business types, FAQPage on FAQ pages that already show questions.
What does the Fix Report check for Structured Data?
- LocalBusiness JSON-LD (local-service sites with verified local presence): present, required fields, recommended fields
- FAQPage quality: if FAQ schema exists, do questions include acceptedAnswer? Do schema questions match visible copy?
- FAQ page expectation: if the page is an FAQ URL with visible questions, is FAQPage present?
- BreadcrumbList: inner pages should declare a trail; items need ordered positions and absolute URLs
- Absolute schema URLs: url, logo, image, and sameAs should be http(s), not relative paths
- Page-type expectations: LocalBusiness on a local homepage; Organization, WebSite, and SoftwareApplication on SaaS; Article or BlogPosting on editorial pages for SaaS, ecommerce, content, and other, not for local_service
Why does malformed JSON-LD cost more than missing markup?
Google's structured-data intro is explicit: markup should match the visible page. Assistants that fetch HTML can read JSON-LD in the same response. On local-service sites, when LocalBusiness JSON-LD confirms the name and phone a person can see, the machine has a verified entity. When it contradicts those facts, the machine has a reason to distrust the source. That name-and-phone consistency check does not run for SaaS or other non-local types.
Missing optional FAQ schema on a homepage is not scored as a hole, the FAQPage rule returns no data when the homepage simply has no FAQ markup. Broken FAQ schema is scored. Schema-only questions that never appear in the HTML are scored. That is the honesty rule: we do not punish you for skipping a type the page does not need, and we do not reward markup that claims hidden content. Product schema is not a rule in this pillar. Do not expect a Product finding.
What does contradicting LocalBusiness schema look like in practice?
Imagine a clinic. The visible H1 and title use the current practice name. The footer shows a phone number. The JSON-LD still uses last year's DBA and a disconnected number. Required LocalBusiness fields are present, name, address, telephone, url, so a rich-results tester might look "complete." The visible-consistency rule still flags the mismatch: schema name does not appear in the title or H1, and the telephone digits do not match a visible phone. Neither site is a named customer. Neither example is a live result.
A second clinic has no LocalBusiness JSON-LD at all. For a local-service site with verified local presence, that is a high finding. The Fix Report would draft LocalBusiness markup from facts already on the page, name, address, phone, url, and omit fields it cannot verify, such as a geo coordinate or a price range the copy never stated.
The pattern is the one the product is built to catch. Complete-looking schema that disagrees with the page is worse than a missing optional type. Illustrative example, not a real business or report result.
Which schema types does the scoring engine actually check?
LocalBusiness runs for local-service sites when the audit can verify local presence. Missing LocalBusiness is a high finding. Incomplete required fields, name, address, telephone, url, are high or medium depending on how many are missing. Missing recommended fields such as areaServed, openingHoursSpecification, sameAs, priceRange, description, or geo are a lighter finding. If local presence is not verified, the rule is not scored.
SaaS homepages are expected to declare Organization and WebSite. Homepage, feature, and pricing URLs are expected to declare SoftwareApplication. Editorial pages, inferred as article, including blog posts, are expected to declare Article or BlogPosting on SaaS, ecommerce, content, and other sites, not on local-service sites. FAQ URLs with visible questions are expected to declare FAQPage. BreadcrumbList is expected on non-home pages, with a stronger push on service, location, and article URLs.
That list is the product roster. It is not "every schema.org type." There is no Product rule in this pillar. There is no AggregateRating rule. The Fix Report will not draft review stars or a Product offer to fill a gap we do not score.
When is missing FAQ schema a finding, and when is it skipped?
If the homepage has no FAQPage JSON-LD, the FAQ-schema quality rule does not treat that as a deficiency. Optional markup that is absent is not a penalty. If FAQPage is present and some questions lack acceptedAnswer, that is a malformed-questions finding. If FAQPage is present and the questions or answers do not match visible copy, schema-only questions, question mismatches, answer mismatches, that is a consistency finding.
A dedicated FAQ page with visible questions and no FAQPage JSON-LD is a medium finding. Put the answers in HTML first. Then add markup that repeats those strings. Google has limited FAQ rich results in classic Search; that does not make matching FAQPage useless for machines that parse JSON-LD, and it does not make mismatched FAQPage safe.
The Fix Report drafts FAQPage JSON-LD only when visible FAQ copy exists and the facts are safe to mark up. The draft is labeled. Human review before publish. Schema that claims answers the page does not show is the failure mode this pillar is built to catch.
Why do relative URLs and broken breadcrumbs fail?
JSON-LD url, logo, image, and sameAs fields should be absolute http(s) URLs. A logo path of "/logo.png" is a string a remote crawler cannot resolve without guessing the origin. The URL rule flags those relative values. The evidence lists the entity type and the field.
BreadcrumbList on an inner page should list ListItem entries with positions starting at 1 and item URLs that are absolute. Missing BreadcrumbList on a non-home page is a finding, medium on service, location, and article pages, low on other inner pages. Invalid positions or relative item URLs are a medium finding even when the type is present.
These checks run on the pages in the audit. They do not require a named customer, a ranking promise, or a live citation from ChatGPT. They require the machine-readable layer to be fetchable and internally consistent.
What will the Fix Report draft, and what will it refuse?
When visible facts support it, the report includes JSON-LD snippets for the types in scope: LocalBusiness, Organization, WebSite, SoftwareApplication, FAQPage, BreadcrumbList, Article. Unsafe or unverified fields are omitted, invented geo, invented hours, invented ratings, invented sameAs profiles. The draft is a starting point labeled as suggested by Answer Patch AI.
If the page never states a phone number, LocalBusiness telephone stays out. If there are no visible FAQ answers, FAQPage stays out. If you want a type we do not score, do not expect a finding or a snippet. Product is the usual example.
Start with a free homepage score if you want a teaser first. Schema snippets that match the page are what you buy. We improve the signals assistants are known to weigh. We do not control citations, rankings, or traffic.
What might a Structured Data finding look like?
- High
LocalBusiness schema is missing
- Medium
Schema facts do not match visible page facts
- Medium
FAQ schema answers do not match visible answers
- Medium
BreadcrumbList schema is missing
Illustrative examples, not from a real scan.
Where do these claims come from?
Claims on this page cite the sources below. Numbers that are not in those sources are omitted.
Frequently asked questions
Do I need structured data (JSON-LD) for AI search?
JSON-LD lets a machine verify entities and facts instead of inferring them from prose. The Fix Report's Structured Data pillar checks the types the scoring engine actually validates, LocalBusiness, FAQPage, BreadcrumbList, Organization, WebSite, SoftwareApplication, and Article where they apply, and whether that markup matches visible copy.
Which schema types does Answer Patch check for?
LocalBusiness for qualifying local-service sites; FAQPage quality and visible-copy match when FAQ schema is present, plus FAQPage on FAQ pages that already show questions; BreadcrumbList on inner pages; absolute URL fields; Organization, WebSite, and SoftwareApplication on SaaS surfaces; Article or BlogPosting on editorial pages for SaaS, ecommerce, content, and other, not local_service. Product schema is not a scored rule.
Is broken schema worse than missing schema?
Malformed FAQ schema, schema that contradicts visible name or phone on local-service sites, and schema-only questions that never appear in the HTML are scored as problems. Missing optional FAQ markup on a homepage is not treated as a deficiency. Missing LocalBusiness on a qualifying local homepage is a high finding. Missing schema is a missed confirmation; contradicting schema is a trust problem.
Other audit pillars
Read more on the blog
Also see: Structured Data
Get JSON-LD snippets in a Fix Report
Start with a free homepage score. No account. The $19 Fix Report includes schema snippets when the visible facts allow safe markup.