HTML-in-Canvas: Google brings the DOM inside WebGL and WebGPU — and the web actually goes 3D

The HTML-in-Canvas API origin trial is live. It lets developers integrate real DOM elements — text, forms, images — into WebGL and WebGPU scenes while keeping them accessible, indexable and translatable. A small change with outsized implications.

The problem it solves

Until now, building a 3D experience in the browser meant giving up almost everything the DOM takes for granted: selectable text, working forms, screen readers, the browser's built-in translation, Ctrl+F search. A WebGL scene was a painting on a canvas — not something the browser could read.

What changes

With HTML-in-Canvas, real DOM elements can be integrated directly into WebGL and WebGPU scenes. A headline inside a 3D scene remains an h1 to the browser: indexable by Google, readable by a screen reader, translatable by the system, selectable by the mouse. Jake Archibald, in a demo posted on X, showed text selection working on a curved 3D surface by simply moving the underlying DOM element beneath the pointer.

Who it's for

The API is in origin trial — so experimental — but Google's message is explicit: immersion should not cost accessibility. It speaks to anyone building e-commerce with 3D configurators, in-browser video editors, interactive portfolios, complex visual dashboards. All contexts where, until now, you had to choose between "beautiful" and "functional."

Inside the agentic-web push

The announcement is part of the 15 updates Google brought to I/O 2026 on the agentic-web theme. It's consistent with WebMCP: a web that's richer visually but also more structured semantically, where AI agents and assistive tech can actually read what's on screen rather than guess from pixels.

← Back to all announcements