Definition
Plain language
The structured map of elements inside a web page that browsers and AI agents read instead of looking at raw pixels.
As stated in the literature
Document Object Model — the tree representation of HTML elements exposed by a browser, used by web-browsing agents and verifiers as a non-visual source of page content.
Why it matters: For web agents, working from the DOM is faster, more reliable, and far cheaper than parsing pixels — but it also means they can be misled by hidden DOM content users never see.
For example, a browsing agent reads the page's DOM to find that a button labeled 'Buy now' has id='checkout-btn' and clicks it directly, bypassing image recognition.
Heard on the show
“A DOM parser that pulls actual HTML elements out of a page.”Episode 062 — Treating Hallucinations as Exploits: A Gate-Based Architecture for Agent Safety