Site Behavior Lab
sitebehavior.org
- Web
- Self-hosted
6.2 Club score Visit sitebehavior.org
What the maker publishes
- Accounts and data sharing
- The reference site has no accounts, sign-ups, or passwords, and it does not sell, rent, or share scan data with data brokers. source
- Blocking simulation
- The Brave Shields comparison simulates list blocking in the scanner and is not a live Brave-browser visit. source
- Brave-list simulation
- One comparison is a normal visit against a visit that actively blocks with Brave's ad-block engine and default Shields lists, and it is not a live Brave-browser visit. source
- Comparison modes
- Comparison modes run two visits, and consent mode tries to click Accept all on one visit and Reject all on the other. source
- Corpus refresh
- Featured-site rescans are disabled as of 2026-11-14, so the committed corpus is not being refreshed. source
- Data minimization
- Scan submissions strip query strings and fragments, and stored reports omit them; reports are not linked to user identity or IP address. source
- Detector accuracy
- No detector in this release has a published precision or recall. source
- Founded
- 2026 source
- Intended audience
- The project is for people seeking checkable evidence about website behavior, including researchers using its dated, versioned corpus. source
- Not a legal verdict
- A report is investigative evidence that requires independent corroboration and is not a standalone legal determination. source
- Open source
- The scanner, catalogs, detectors, and site are open source under AGPL-3.0-or-later. source
- Public scan quota
- The public instance allows 6 scan tokens per minute and 120 per day per client, and a comparison or after-scrolling check costs two tokens. source
- Purpose
- Site Behavior Lab visits public websites with a controlled browser, records bounded evidence, and publishes reports rather than ratings. source
- Recorded evidence
- Evidence may include network requests, catalog labels, privacy-filtered cookie and storage snapshots, browser API signals, ad-pixel events, DNS CNAME findings, and policy cross-checks. source
- Recording cap
- Each visit records at most 1,000 requests, and an after-scrolling check may send up to 500 more that are not saved. source
- Report limits
- Results describe bounded visits and may be incomplete; counts are lower bounds, and some browser activity is unsupported or unrecorded. source
- Retention
- The privacy page states reference-deployment reports are automatically deleted after about seven days, configurable by instance operators. source
- Scanner
- The scanner runs a controlled headless Chromium visit and records requests, cookies, service-catalog matches, and browser signals. source
- Security contact
- The project is volunteer-maintained and aims to acknowledge vulnerability reports within a few days. source
- Security reporting
- The project asks users to report vulnerabilities privately through GitHub's private reporting channel. source
- Self-hosting
- The supported scanner is a Node/Playwright container built from the project's Dockerfile. source
- Service catalog
- Report labels come from 146 maintainer-reviewed domain mappings that are maintained in the repository and do not import a competitor tracker dataset. source
- Share retention
- The public instance keeps saved reports for seven days or 500 reports, whichever prunes first. source
- Support and corrections
- Users can report incorrect reports for correction through an append-only corrections ledger. source
Checked against the maker's pages on 14 Nov 2026.