Schema validation errors, what breaks and what you can ignore

A Series B consumer brand came to me last month with 247 schema validation errors flagged in Search Console. The marketing lead wanted to know which ones mattered. I asked her to show me the pages. Rich results were live. ChatGPT was citing the product pages. Perplexity was pulling the FAQ content.

I told her to fix three errors and ignore the rest.

That is the honest story of schema validation for a small team. You will see errors. Some of them break things. Most of them do not. The validation tools flag everything, and you do not have time to chase everything. You need a filter.

Here is mine, structured as the questions I hear in almost every schema audit.

Do I need to fix every error Search Console shows me?

No. Search Console flags warnings and errors. Warnings do not block indexing but may reduce visibility or rich result eligibility, so I fix warnings only if they touch schema types that matter to the client's AI search strategy. Errors I fix selectively.

The question is not "is this technically invalid," the question is "does this schema type do work for us." If you have 40 errors on FAQPage markup and your site is a commercial ecommerce brand, you're shipping markup that will never render as a rich result because FAQ rich results now only appear for well-known, authoritative government and health websites. Delete the markup. The error count drops to zero and nothing of value was lost.

The schema types that still work in 2026 are Product, Article, Recipe, Review, Video, Event, JobPosting, LocalBusiness, BreadcrumbList, Organization, and WebSite. If your errors are in those types, fix them. If your errors are in deprecated types, remove the markup.

What actually breaks rich results and AI citations?

Three things break eligibility more than anything else I see.

Missing required properties. For article schema, this often means no image, author, or datePublished. For Product schema, it is usually missing name or image. Google will not show a rich result if a required property is absent. AI engines are less strict, they will sometimes cite a page with incomplete schema, but the citation rate is measurably lower in the audits I run. Fix required properties first.

Wrong date format. Dates in schema have to follow a specific format (ISO 8601, the 2026-04-06 style). I see clients shipping human-readable dates like "April 6, 2026" or "06/04/2026" inside `datePublished` or `dateModified` fields. We find sites feeding in dates in a human-readable format instead, which quietly invalidates the markup. The fix is straightforward. Use `YYYY-MM-DD` or `YYYY-MM-DDTHH:MM:SSZ` if you need a timestamp.

Schema that does not match visible content. The schema has to describe what a visitor actually sees on that page, not what the business wishes were there. If your Product schema says "InStock" but the page says "Out of Stock," Google reads that as an attempt to game rich results and can strip them sitewide. If your AggregateRating shows 4.8 stars but the page has no reviews, same problem. Schema is not aspirational. It is descriptive.

How do I validate schema without wasting time?

I use two tools and I run them in sequence.

Google Rich Results Test tells me whether a page is eligible for rich results in Google Search. If it passes here, the markup works for Google. If it fails, the tool tells me which required property is missing or which syntax error broke eligibility.

Schema Markup Validator is the official schema.org tool. Today, Google provides two distinct tools, the Rich Results Test (RRT) and the Schema Markup Validator (SMV). I use Rich Results Test first because it answers the commercial question (will this page get a rich result). I use Schema Markup Validator second because it catches structural issues Rich Results Test sometimes misses and because it validates against schema.org rules, not just Google's subset. That second pass matters for AI engines, which read schema.org types Google does not care about.

I do not validate every page. I validate one template per schema type. If the Product template passes, I assume all Product pages pass unless Search Console flags a specific URL. If Article markup fails, I fix the template and redeploy. Validation is a spot-check, not a site-wide ritual.

What about syntax errors, missing commas and broken JSON?

Frequent mistakes include missing commas, quotation marks, or brackets, along with misnested properties, any of which can cause parsing errors. These break everything. If your JSON-LD has a syntax error, the entire block of schema is ignored by Google and by AI engines.

The good news is that syntax errors are binary. They either exist or they do not. Run your JSON-LD through Rich Results Test or paste it into a JSON validator. If it parses, you are done. If it does not parse, the validator will tell you which line has the problem.

I see syntax errors most often when clients hand-edit schema in a CMS without a linter. The fix is to move schema generation into the template layer or use a plugin that outputs valid JSON-LD by default. Do not ask a content editor to write JSON by hand.

Should I keep FAQPage and HowTo markup if it still validates?

No. FAQPage and HowTo rich results were effectively retired for the general web back in August 2023. The markup still validates because the schema.org types still exist, but the markup is now dead code. Google will not render it as a rich result. AI engines may still read FAQPage content, I have seen Perplexity cite FAQ sections, but the markup itself does not improve citation rates compared to a clean HTML FAQ without schema.

Remove it. The validation error count drops and the page weight drops and nothing breaks.

What happens if I ignore errors that do not block rich results?

Nothing breaks immediately. But schema that ships with errors and never gets fixed becomes stale. Schema implemented once and never updated is worse than no schema because it actively misleads crawlers and AI systems. A Product page that says "InStock" six months after the product sold out is worse than no Offer schema at all.

The operational rule I give clients is this. Fix errors on schema types that do work for your AI search strategy (Product, Article, LocalBusiness, Organization). Ignore or remove errors on deprecated types. Run a quarterly schema audit to catch staleness (prices that changed, availability that flipped, authors who left, events that ended). If you do not have time for quarterly, make it annual. If you do not have time for annual, you probably should not be shipping schema at all.

I have worked with founders who spend three weeks chasing every warning in Search Console and ship nothing. I have worked with founders who deploy schema once, ignore every error, and then wonder why rich results disappeared eight months later. Neither approach works.

The middle path is knowing which errors break things and which errors are noise, then acting accordingly. If you are running schema audits as part of a broader AI search rollout and want structured feedback on what actually matters for your vertical, our consultancy work includes schema review and template-level fixes that stick.