[Poster] GTM-Config: Reverse Engineering Google Tag Manager in the Wild

Presented at ACM IMC 2026, 2026

Poster at the ACM Internet Measurement Conference (IMC) 2026 in Karlsruhe, Germany, with Eman Nabeel and Zubair Shafiq.

Google Tag Manager (GTM) is the most widely deployed tag management system, observed on 54% of the top 1M websites. GTM compiles a publisher’s entire tracking configuration (the pixels to load, the events each pixel tracks, and the rules that determine when those events fire) into a single JavaScript file, gtm.js, served to every visitor. Prior work on GTM has primarily focused on which tags publishers deploy and what traffic those tags generate, with limited attention to the configuration itself. Dynamic analysis recovers it only partially.

We present GTM-Config, a static analysis framework that decodes a website’s complete tracking configuration (pixels, events, triggers, and consent gating) from its gtm.js container. We build GTM-Config by iteratively configuring seven major tracking pixels in our own GTM deployment and diffing the compiled containers. We then crawl 32K general, health, and children’s websites; 7,277 (22.7%) of them load GTM, and we decode their configurations, comprising roughly 200K tags. Our preliminary analysis shows that 87.1% of websites fire at least one tracking event by default on page load, that only 6.2% consent-gate any event, and that consent-gating mechanisms appear on only 2.9% of health websites, compared to 30.3% of general websites.

Download Poster