1.5% of JSON-LD Blocks on Top-1M Sites Fail to Parse, Breaking Schema on 2.2% of Pages That Use It

Summary

Picture a home builder's developer pasting the company description from a brochure into the site's schema markup. The text has a line break in the middle. The page looks fine, the markup looks fine to the eye, and a strict parser that reads it throws the whole block away because it is not valid JSON.

JSON-LD is unforgiving. One stray character can invalidate everything in a script block, including the Organization, LocalBusiness or Product details the site meant to share. There is no partial credit.

We parsed 1,099,743 JSON-LD blocks from top-1M pages exactly as they were served. 1.5% failed, which means about 1 in 68 blocks says nothing to a strict parser.

What we measured

We used the October 2026 HTTP Archive crawl (mobile) of top-1M sites, which captures each site's homepage plus one inner page. For every page we took the raw JSON-LD script contents from the served HTML and ran each block through a strict JSON parser in BigQuery. We recorded whether it parsed and, if not, the parser's error message, which we grouped into causes.

Results for the rendered page (after JavaScript runs) are almost identical: 1.4% of 1,183,013 rendered blocks fail. Platform figures are grouped into three categories (open-source CMS, hosted builder or commerce platform, custom or other stack) based on HTTP Archive technology detection.

About one block in 68 is broken

Most sites get it right. The ones that do not lose the whole block, not just the faulty line.

The top 1,000 sites are the most careful. Below that, the failure rate is flat at around 1.5%.

A pasted line break is the top cause

We grouped parser error messages into causes.

Errors like these usually trace back to text pasted into markup or inserted by a template without escaping, though the crawl cannot show how each block was made. Comment or CDATA wrappers break a strict parser but may be tolerated by some search engines.

Failure rates by type of stack

Grouped by category of platform detected on the page.

Custom or other stacks fail at about twice the rate of open-source CMS sites.

What this means

Run your key pages through a schema validator or a JSON linter and fix anything it rejects. Escape line breaks in text values as \n, put multiple objects in one array or an @graph, remove trailing commas, and do not wrap JSON-LD in comments or CDATA. If your markup is built from database fields, make sure the template escapes them. Template escaping is a frequent source of these errors, and good answer engine optimization services fix it at the template level so it stays fixed.

A broken block hides everything in it from machines, including the facts AI engines use to describe your business. Our free AI visibility audit includes a check for markup that fails to parse, a check any GEO agency should run before writing new schema.

Limits

We used a strict JSON parser. Some search engines may recover from certain errors, such as comment wrappers, so the share that is truly lost to them may be lower.

HTTP Archive crawls one homepage and one inner page per site.

Error grouping relies on parser messages and is approximate; 7.0% of failures fall into Other syntax error.

Stack categories depend on HTTP Archive technology detection.

Frequently asked questions

Related studies

• Niche schema types AI answers can readOnly 24.5% of Car Dealer Sites Mark Up Their Vehicles
• Niche schema types AI answers can read28% of Recipe Pages Leave Cook Time Out of Their Markup
• Niche schema types AI answers can read98.8% of Sites Using Hiring Software Lack Job Markup
• Niche schema types AI answers can read86% of Sites With Event Tools Show No Event Markup