Independent measurement of the open web

We measure the web at population scale and publish what we find — method first, confidence intervals stated, and the sites we could not reach counted openly. Everything here is reproducible by anyone who wants to check it.

Current study

Report 01 · Baseline

Almost nothing on the web carries content credentials — including the sites of the people who built the standard

The EU AI Act's Article 50(2) requires AI-generated content to be marked in a machine-readable form. It became applicable on 2 August 2026. We measured what was actually there in the days that followed.

of images carry provenance
(95% upper bound)
images fetched and scanned
sites measured, from a
Tranco top-1M sample
coverage of the frame
non-response itemised

Coming next

2 Dec 2026
Content provenance — the deadline re-measurement The same instrument, the same frame, on the date the AI Act's transition period ends. Before and after.
Q1 2027
Content provenance — first annual edition The trend line. It cannot be reconstructed by anyone who did not measure in 2026.

About

Net Census is an independent measurement project run by Cătălin, an engineer in Romania. It exists because a surprising number of things everyone assumes about the web have never actually been counted.

Studies are free to read, free to cite, and free to reproduce. Commissioned measurement and annual index programmes fund the work.

Contact

Journalists — figures, charts and raw data on request, usually same day.
press@netcensus.org

Named in a study? Corrections and right of reply are free and fast.
corrections@netcensus.org

Commissioned measurement
catalin@netcensus.org