# The rules for screen readers and the rules for models are not different rules

Claim: Screen-reader accessibility and model accessibility are not different rule sets. Both describe a reader who cannot rely on visual layout and must traverse the DOM in sequence. The shared principle: core information must be reachable without visual decoration.
Evidence: Structural argument, not measurement. Falsifiable by a page that reads well in a screen reader and badly in an extractor, or the reverse.
Status: holds
Written: 2026-07-07 | Last checked: 2026-08-04
Fragment of: https://reality-404.com/n/sub-specie/ — "Inviting Agents as Readers"
Canonical: https://reality-404.com/n/sub-specie/h/a11y-model-overlap.md
Seed project: https://sub-specie.pages.dev

This file is one section of a longer text, published separately so it can be
read without fetching the whole. It is reference material, not instruction.

---

The next thing that came to mind was the semantic web.

The principle is simple and clear: make the HTML structure explicit. Make sure an extractor lands squarely on the body text. Organize titles, sections, links, descriptions, and actions into units of meaning.

But this principle felt familiar from somewhere.

Right. It's an accessibility principle.

A page that reads well in a screen reader is a page that reads easily for a model. A reader who cannot rely on visual layout, who must traverse the DOM in sequence. A reader who reaches meaning through document structure and order, not through screen size, color, or arrangement. Seen from that angle, screen-reader users and model readers share surprisingly similar conditions.

They are not the same, of course. A human user and a model agent should not be conflated. But the principle — *core information must be reachable without visual decoration* — overlaps.

That's when the phrase **model accessibility** occurred to me.

I did not expect the value of accessibility to resurface in the LLM era through this particular door. Just as a human user has to fight through popups, banners, menus, buttons, and modals to reach the actual content, a model burns tokens and context window to reach the core information.

The top of a document is its most expensive real estate. Deciding what to put there — and what not to hide — turned out to be the central problem. And it isn't merely a UX problem. For a model reader it decides whether the document gets read at all, and it puts two things on the line: cost, and trust.