A14y Agent Readability

a14y scores any web page — or an entire site — for how readable it is by AI agents. When ChatGPT, Claude, Copilot, or Cursor fetch the web on a user's behalf, most of what makes a page work for humans (layout chrome, interactive widgets, client-side routing) actively gets in the way. The a14y scorecard measures the signals that actually matter to an agent: whether your content is discoverable, parseable, and comprehensible.

What it does
Open the extension on any page and hit **Check this page** to run the scorecard against that single URL, or **Scan whole site** to crawl from the entry URL following same-origin links. In seconds you get a 0–100 score plus a per-check breakdown: which signals you're nailing, which ones agents can't find, and a link to the fix guidance for each failing check.

Every audit runs entirely in your browser. No server, no account, no data collection — just the extension and the page you're auditing.

What it measures
- Discoverability — robots.txt, sitemap, llms.txt, canonical
links, redirect chains, HTTP 200 hygiene.
- Parsing — content-type headers, HTML lang attribute, meta
description, OG tags, JSON-LD, Markdown alternate links, text ratio.
- Comprehension — headings, code-language tags, semantic
structure, and a growing set of content-shape checks.

Each check is versioned in an open scorecard at a14y.dev — a pinned v0.2.0 today — so scores from different runs stay comparable over time.

Who it's for
- Docs teams shipping content that AI coding assistants retrieve and cite.
- Marketing and product pages whose biggest source of referrals is about to be agents, not search.
- Engineers debugging why their content shows up wrong in LLM answers.
- Anyone who wants to run the same open-source audit a14y.dev runs from the CLI, but without leaving the browser.

Privacy
Every audit runs in your browser — the URLs you enter and the pages the crawler fetches stay on your machine. a14y collects anonymous usage data by default (audit started/completed events with bucketed score and duration; never URLs, page content, or per-check details) to help us improve the extension. You can turn it off in the extension's options. Full policy: https://a14y.dev/privacy/

Open source
Spec, scorecard, and tools live at https://github.com/timothyjordan/a14y. The same scoring engine powers the `a14y` CLI on npm and this extension — same URL + same scorecard version = same score, every time.