menu
Check my site →

CLICK CODED · raw benchmark data

The files behind the correction notes

Every benchmark page on this site carries a dated note when a number moved on re-measure. This page is where the numbers actually came from. No summary, no chart, just the files a re-measure produced, so you can check the claim yourself instead of trusting the note.

The files

FileWhat it isSize
rebenchmark-results-v1_1.jsonlThe 2026-07-31 re-measure: one JSON line per site, 231 sites, every check's id, weight, pass/fail and detail string exactly as the live engine (methodology v1.1) produced it. This is the source for the "re-confirmed 2026-07-31" notes on the airline, telecom, e-commerce and government benchmark pages.1.4 MB
rebenchmark-corpus-v1_1.csvThe 231-site list itself: sector, company name, domain, and which source benchmark page each site came from. What was actually tested, not just the results.14 KB
accuracy-audit-remeasure-2026-07-30.jsonlAn earlier, 168-site re-measure run one day before the file above, during the 2026-07-30 accuracy audit. Named-company spot-checks from that pass (Northwestern Mutual, Charles Schwab, the airline and government re-measures) are cited directly in corrections.1.1 MB
newsroom-robots-2026-07-30.jsonA one-off robots.txt snapshot across major news domains: which AI crawlers each one names and blocks by name, and whether a wildcard rule blocks everything else. Not a scored benchmark, a raw policy read.12 KB

All four are plain JSON/JSONL/CSV, no login, no tracking pixel. Open them in a text editor or load the JSONL files as newline-delimited JSON in any language. Each check record uses the same shape documented on the methodology page: id, name, weight, pass, and a plain-English detail string.

What isn't here

This is raw re-measure output, not a maintained public dataset. The packaged, documented dataset (20 major SaaS homepages, one row per site, a real column schema, sold as a product) is the AI Visibility Benchmark Dataset. These files are what fed the corrections. That product is what's built to be queried.

Found a number that doesn't match these files

Email [email protected] with the page and the line. Every confirmed discrepancy gets a dated correction on the corrections page, not a silent edit.

Published 2026-07-31. Built by Click Coded, AI-operated and human-reviewed. Contact: [email protected]

More from Click Coded: AI visibility checker · methodology · State of AI Visibility 2026 · corrections.

Want to hear when we ship something new?

We add new free tools and benchmarks periodically. Get an email when we do, no spam, unsubscribe anytime.