Most schema shipped in 2026 gets validated against two tools: Google’s Rich Results test and Schema.org’s structured data validator. Both are useful. Both miss issues that matter for AI citation.
The gap between “the schema passes validation” and “the schema actually works for AI search optimization” is wider than most agencies realize. Here are the validators worth using, what each one catches, what they all miss, and what to add to a complete schema audit workflow.
Three primary validators in common use plus three more that fill specific gaps.
Google’s Rich Results test catches errors that would prevent Google from displaying your content as a rich result. Missing required fields for Article, FAQ, Recipe, Event, and other supported types. Invalid date formats. Broken schema structure. The validator also previews what the rich result would look like.
This is the right starting validator. It catches the issues that affect Google’s display. If the Rich Results test fails, fix the failure first before anything else.
Schema.org’s structured data validator catches type-level errors. Invalid property names, wrong value types, missing context. It validates against the Schema.org vocabulary directly without applying Google-specific rules.
This is the right second validator. It catches issues that are technically wrong even if Google would accept them. The downside is that it doesn’t tell you what Google or AI engines will actually do with the schema.
Bing’s Webmaster Tools markup validator is the third commonly-used tool. It validates against Bing’s understanding of structured data, which differs from Google’s in small but real ways.
Bing’s tool is worth running specifically because ChatGPT browse and Microsoft Copilot both use Bing’s index. Schema that works for Google but breaks for Bing affects two of the most important AI engines.
Missing or weak @id linkages between schema blocks. A page might have valid Organization schema, valid Article schema, and valid Person schema. Each passes validation independently. But if the schemas don’t reference each other via @id, AI engines don’t understand the relationships. The brand-author-content graph is missing.
This is one of the highest-impact issues for AI citation. None of the standard validators catch it because the schemas individually validate. The issue only shows up when an AI engine tries to reason about the relationships.
Schema that contradicts visible page content. Schema says the price is $489 a month. The page body says “starting at $397.” Both schemas pass validation. The contradiction confuses AI engines, which often de-weight both signals.
The standard validators check syntax, not consistency with the surrounding content. A consistency check requires a separate audit step.
Outdated dateModified relative to content changes. The page has been updated three times in the last month. The dateModified is still set to the original publish date. Google and AI engines that weight recency see a stale page when the page is actually current.
The validators don’t know when the surrounding content changed. The dateModified discipline is a workflow issue, not a validation issue.
Speakable schema pointing to non-existent CSS selectors. The schema specifies cssSelector: [".speakable-answer"]. The page doesn’t have any element with class="speakable-answer". The schema validates. Voice assistants find nothing to read.
The validators check schema structure, not whether the referenced CSS selectors actually exist on the page.
Schema with sameAs URLs that 404 or redirect. Your Organization schema’s sameAs includes your old LinkedIn URL that now redirects. The schema validates. The verification path for AI engines is broken.
The validators don’t fetch the URLs to check whether they resolve correctly.
AggregateRating data that doesn’t match third-party platforms. The schema says you have a 4.7 rating with 183 reviews. Your G2 page says you have a 4.3 rating with 47 reviews. The schema validates. AI engines that cross-reference detect the discrepancy and lower trust in the source.
The validators don’t cross-reference external platforms.
Relationship audit via @id graph. Visualize the schema graph for your pages. Organization should bind to Person (founders) via founder. Person should bind to Organization via worksFor. Article should bind to Person via author and to Organization via publisher. The graph should be connected.
A simple JSON-LD parser plus a graph visualization tool makes this an audit step. Schemas that exist on the page but aren’t connected to the graph need to be linked.
Content consistency audit. Read the page. Compare every claim in the schema to the visible content. Price, hours, ratings, descriptions, addresses. Flag mismatches. Fix the underlying source (usually the visible content needs to be updated, sometimes the schema).
This is manual work that doesn’t scale to thousands of pages. For brand-critical pages (homepage, pricing, key landing pages), it should happen quarterly.
External link verification. Fetch every URL in sameAs fields, url fields, and worksFor references. Confirm each one resolves to a 200 with content that matches the schema’s claim. Fix or remove broken links.
This can be automated with a script that runs weekly.
CSS selector existence check. For speakable schema and any schema that references DOM selectors, verify the selectors actually match elements on the page. Use a headless browser or DOM-aware validator.
This is also scriptable.
Third-party platform cross-reference. For aggregateRating, fetch your G2, Capterra, Crozdesk, or Trustpilot pages and confirm the numbers match. If they don’t, update the schema to match third-party reality.
This requires either manual review or a custom integration with third-party APIs.
We run a four-validator workflow on every schema deployment:
The custom script is roughly 200 lines of Python. It parses the JSON-LD from a URL, builds the relationship graph, fetches sameAs links, checks CSS selectors against a rendered DOM, and cross-references aggregateRating data where available.
The script isn’t public yet but the logic is straightforward to replicate. If you ship schema professionally, building a similar audit tool is a one-day investment.
Three cases where you can skip the custom workflow.
Small brand with simple schema needs. A solo professional with one page of Person schema and one Article schema per blog post doesn’t need the full custom audit. The standard validators catch the relevant issues.
High-confidence content with stable assertions. A brand whose schema claims rarely change (the address, the founder, the company name) doesn’t need quarterly consistency audits. The annual review is enough.
Low-stakes content. Schema on personal blogs, internal documentation, or content that doesn’t need AI citation traffic doesn’t justify the full workflow. The standard validators are the right cost-benefit balance.
For commercial B2B sites where AI citation matters and the schema is doing real work, the full workflow pays off.
If you’re hiring an agency for schema or AI search optimization work, ask one question: “what does your schema validation process look like beyond the Google Rich Results test.”
The agencies that have a real workflow will describe their relationship audit, consistency check, and external link verification process. The agencies that just run the Google test will say “we use Google’s tools.”
The gap between the two answers maps to the gap in citation outcomes. Schema that passes Google’s validation is the floor. Schema that passes a complete audit workflow is what actually performs in AI search.
If you want a schema audit that catches what the standard validators miss for your existing schema, the fit call covers it. The audit ships in week one with prioritized fixes.
Related reading:
– Speakable schema: the underused signal that helps voice assistants find your content
– Schema for service-area businesses: LocalBusiness done right
– Product schema for SaaS: turning your pricing page into a citable answer
– FAQPage schema is the biggest AI citation move you can make
– Schema markup for AI search 2026: field guide
Pricing note: This Field Note includes earlier launch pricing in places. Current NetPageTwo pricing is $489/month for Visibility and $1,297/month for Visibility + Revenue.
