What is actually in tag managers

Published containers, read and counted. Vendors describe what their tools can do; this is what sites actually do with them — and it is the evidence behind every claim we make about GTM.

Everything connects

These are not three lists. A host, a container and a site are three views of one graph, and every record links to the others: from a tracker to the sites it reaches, from a site to everything it lets in, from a container to both. Every view has its own address, so any of it can be linked to directly.

How this is measured

Containers are discovered by crawling a ranked list of domains and fetching each published container from Google. Everything reported comes from that artifact — the same file served to every visitor of the site. No login, no vendor cooperation, nothing private.

The sample is 42,089 containers as of 2026-08-22, from 175,929 ranked domains.

What these numbers are not

Not a sample of the web. These are sites ranked highly enough to appear in a popularity list, which skews commercially active. About 41% of domains we tried could not be read at all — timeouts, blocks, bot walls — so figures describe sites that answer an HTTP request, not the internet.

Not a measure of behaviour. A container naming a host means a tag can reach it, not that it fired for any particular visitor. Configuration and execution are different questions, and only the first is visible in the artifact.

Where a claim needs something the container cannot show — where data actually goes, for instance — we measure it with a real browser instead and say so.