Site Behavior Lab
Checking

See what a site does, not just what it says.

Run a controlled Chromium visit and read what the page actually loaded: request rows, cookie records, service-catalog matches and browser signals from that one visit.

Try
Run

One controlled visit.

GPC comparison is not available from this scanner.

Blocker comparison requires the Node scanner.

One controlled visit that records the requests, cookies, and scripts the page loads, up to a recording cap that is flagged when hit.

Checking public scanner status...

More options
Device
Privacy Signal

Global Privacy Control is a legal “do not sell or share my data” signal sent with every request.

One controlled visit records

Evidence, then interpretation

  • Catalog matchesWhich request rows matched the reviewed service catalog, and what roles does that catalog assign?
  • Third-party cookie recordsWhich cookie records crossed the site's registrable-domain boundary?
  • Named platformsWere requests dispatched to catalogued Google, Meta, TikTok, X, Microsoft, LinkedIn and Pinterest domains?
  • Google Analytics remarketingDid the scan see the Analytics-to-DoubleClick request marker?
  • Fingerprint-like API callsDid canvas, WebGL, or audio behavior cross a documented heuristic threshold?
  • Session-replay signalsDid a catalogued service appear or broad interaction listeners register?
  • Synthetic input checkDid the test value appear in a cross-site request before form submission?

Every report records its own conditions and which evidence families were captured, censored or unsupported. Signals describe one visit, not a verdict about the site.

Public library

98 sites measured, 982 reports on record

Reproducible evidence from controlled visits, not a score. Open a report, or scan a site above.

181News & media
56Health
52Shopping
24Banking & money
median third-party tracking-service requests per site, by category. These categories were measured under different methodology generations, so read each on its own rather than ranking them against each other.
How coverage and category medians are counted

The committed library attempted 104 real sites. 6 sites only produced failed or block-page primary visits, and 5 successfully loaded sites had at least one request-capped recording that remains visible as lower-bound evidence. Category medians use 86 sites with an eligible, request-complete passive lead visit. Each site counts once, even when a comparison loaded both arms. Those sites span 2 methodology cohorts: each category publishes a single cohort, so a median is comparable within a category and not across them, and no one median covers all 86 sites. Requests are also evaluated with the open-source adblock-rustengine and Brave’s default filter lists.

Open report files, PageGraph captures, or comparison tools
Maintainer scan (repository access) (opens in a new tab)

PageGraph imports require a .graphml file and its matching .meta.json sidecar. Browser imports are capped at 8 MB for report JSON, 16 MB for GraphML, and 256 KB for metadata. Unsupported evidence families remain censored rather than guessed.