Get Free Audit

WLDM AI Research Lab

Most agencies publish case studies. We publish the data set.

WLDM runs seven research programmes. This page is where the output lives, not a highlights reel dressed up for a pitch deck: the corpus, the method, the figures and the honest limits, in enough detail that a technical SEO can pull them apart and check every claim against the raw data. The first release measures schema.org deployment across 14,784,536,534 pages of the open web, read over seven contiguous monthly crawls. More will follow from the AI citations work running with DataForSEO, Search Atlas and Surfer SEO. Where a finding is not ours, we say so on the page it appears on. We are not interested in blurring the line between our research and someone else's.

Read the Latest Study

01. Published

Latest paper · Working paper, September 2026

The Shape of Schema.org at Scale: seven months, 14.78 billion web pages

Read the Paper

We processed 700,000 WARC files across seven contiguous monthly crawls, covering 14,784,536,534 HTML pages. A weighted mean of 58.32% of them carry structured data, rising every month from March onward, and 43.5% of what the top 200 types describe is navigation rather than actual content.

Working paper · Corpus CC-MAIN-2026-04 to 2026-30

58.32%

Weighted mean coverage of the 14,784,536,534 HTML pages examined, valid markup or malformed. Coverage climbs from 55.27% in February to 60.05% in July.

127.99B

Individual @type occurrences recovered across the seven months. The top 100 types alone account for 97.87% of them, measured on the June crawl as the representative month.

811,946,273

@type occurrences a strict parser would silently throw away, 0.63% of the total. Property names used as types, casing errors, URLs treated as types. AI training scrapes don't discriminate. They keep all of it.