Open reference data for the web-tracking ecosystem, maintained by webtracking.org. Explorable at data.webtracking.org; measured against 15.7M real pages in the public HTTP Archive crawl for The State of Web Tracking.
Why this exists: the best-known tracker entity graph is licensed CC BY-NC-SA — NonCommercial, ShareAlike — which makes it unusable inside most real products. These datasets are CC BY 4.0: use them commercially, just say where they came from.
| File | What it is | Records |
|---|---|---|
datasets/entities.json |
domain → parent-entity ownership map | 43 entities / 102 domains |
datasets/cookies.json |
cookie-name classification rules (vendor + purpose) | 85 rules |
datasets/ai-crawlers.json |
AI crawler UA tokens with operator + purpose (training / retrieval / search) | 14 |
datasets/ai-referrers.json |
AI-assistant referral signatures (referrer hosts, utm_source tokens) | 11 |
datasets/crosswalks/categories.json |
our category vocabulary ↔ IAB TCF purposes ↔ Tracker Radar (approximate) | 10 |
datasets/queues/*.json |
the curation queues — what to research next, ranked by measured impact | 98 + 35 |
crawl/2026-06/*.csv |
per-chart aggregates from the report (adoption, entities, cookies, AI-robots, consent, CWV…) | 17 files |
Every record carries provenance (how we know) and last_verified (when a human last
checked). status is active | deprecated | merged | disputed.
Once deployed, every dataset is served versioned:
https://data.webtracking.org/v1/latest/entities.json (moving)
https://data.webtracking.org/v1/2026Q3/entities.json (immutable snapshot)
scripts/build-public.mjs regenerates public/ from datasets/; CI fails if they drift.
The queues are generated from the crawl itself, so curation effort goes where reality is, in impact order:
queues/unmapped-domains.json— the top-200 third-party domains by page reach that no entity owns yet. Working top-down grows the entity map in reach order, not alphabetically.queues/unknown-cookies.json— high-volume cookie names no rule classifies (currently attributed only by the entity of their top-setting domain, or not at all). Each needs a researchedcookies.jsonentry with a source.
Every quarterly crawl regenerates the queues; whatever got curated disappears from them.
Corrections and additions are welcome — see CONTRIBUTING.md. The bar is the same one the
report methodology applies to itself: claims carry
sources, and substantiated corrections are applied, dated, and credited.
- Datasets (
datasets/,crawl/): CC BY 4.0 — attribution: “webtracking.org” with a link. - Code (
scripts/,public/index.html): MIT.
See LICENSE.md. Crawl aggregates derive from the public
HTTP Archive dataset. Not affiliated with any vendor named in
the data; names appear for identification.