About and corrections

What this is

Tracker Watch measures, every week, which third-party tracking services news websites contact when a page is loaded with no clicks and no consent given. The method, its limits and the list used to classify services are public on the method page, and every report is published as raw data.

Maintained by jlasenjo.

Corrections and right of reply

Mistakes are possible: a domain can be attributed to the wrong operator, an outlet's own domain can be missed, a page can behave differently for a reader in another country. If you represent a listed site, or you spot an error, tell us:

We re-check what is reported, correct what is wrong, and record each correction in the data and on this page. We do not remove a site from the list at its request, but we will state its reply next to its results.

Independence

No outlet pays for, or is paid by, this project. At the moment it has no income and no affiliate links. If that changes, it will be disclosed on this page before it starts, and it will never influence what is measured or how it is classified.

Reproducibility

Each report records the tool version, browser, locale, time zone, user agent and the place the scan ran from, plus the request log of every pass (without query strings or cookie values). The latest table is available as CSV.

Changes to the method

  • 2026-10-05, filter lists: the tracking domains of the classification list are now also published as two filter lists for ad and tracker blockers (verified entries only, and all entries). They are generated from the same list, change no measurement and are not complete protection.
  • 2026-10-05, what each service did: site pages now show, for each tracking service, the scripts, image and background requests, frames and cookies counted in the browser, and Companies shows how often each operator did each of those things. These figures come from the raw request logs already published, so no earlier measurement changed. A new glossary page explains the categories and what these activities mean.
  • 2026-10-05, two separate tests: “After the banner” presses a consent banner's reject or accept button once and records the next window; “With a blocking list” loads pages while blocking what a public privacy filter list targets. Both are published on their own pages and in their own data folders. The weekly ranking still never interacts with the page, so its figures are unchanged and remain comparable.
  • Schema 0.2, 2026-10-05: reports now record the transfer size of each response, and the site shows the weight of third-party content, the companies reached (new Companies page), a fuller set of downloads (history.csv, entities.csv, index.json) and community-contributed origins, always labelled unverified and never part of the weekly ranking. The scan itself is unchanged: still no clicks and no consent, so earlier counts remain comparable. Reports made before this date have no sizes.
  • 2026-10-07, later: three more entries added to the classification list, marked as unverified: mrf.io and newsroom.bi (Marfeel, audience analytics with a persistent user identifier, according to the company's own documentation) and pbstck.com (Pubstack, ad monetisation analytics and ad management for publishers). They were noticed in a real browser on a Spanish news site. The measurements of 2026-10-04 were recomputed from the same raw logs (nothing measured again): five sites measured from the US gained one or two services, none changed band, and the figures measured from Spain did not change.
  • 2026-10-07: two entries added to the classification list, both marked as unverified: opti-digital.com (Opti Digital) and a-mo.net (AdaptMX), advertising services seen in a real browser as the ad frames through which most other services were contacted on a Spanish news site. The measurements of 2026-10-04 were recomputed from the same raw logs (nothing measured again): seven sites measured from the US gained one service (a-mo.net), none changed band, and the figures measured from Spain did not change.
  • 2026-10-04, same version: the classification list grew from 60 to 147 entries (mostly advertising and identity services) and the band ranges were revised from A 0–2, B 3–5, C 6–9, D 10–14, E 15+ to A 0–2, B 3–9, C 10–24, D 25–49, E 50+, because the larger list moved many counts far above the old top band. All earlier figures were recomputed from the same raw measurements; nothing was measured again. The new entries identify vendors from general knowledge and are marked as unverified on the method page.
  • Version 0.1.0 (first public version): counts of third-party tracking services before any interaction, three passes, median, fixed bands A–E, small hand-built classification list.