LivingPageAudit

How to stop it

Put this in your robots.txt. We read it before anything else and obey it, including Crawl-delay.

User-agent: LivingPageAudit
Disallow: /

A firewall rule blocking the user agent works too, and we will tell whoever ran the scan that your site declined. We do not retry around a block, and we never disguise ourselves as a browser to get past one.

What it does

It runs only when somebody asks for a report on a website, fromlivingpage.app/audit. It is not a background crawler and it indexes nothing: there is no scheduled crawl, and no visit that nobody requested.

• One visit per request. A scan stops after two minutes, 850 requests, or 300 pages — whichever comes first.
• A few pages at a time. Four concurrent requests at most, and one at a time the moment your robots.txt asks us to slow down or your server answers 429.
• PDFs are not downloaded. We send a HEAD and two small range requests — a few kilobytes from each end of a file — to read its size and whether it is tagged for screen readers. The document stays on your server. A reader may ask for one specific file to be examined in full; that is the only time we fetch a whole document.
• Nothing of yours is stored. We keep the report — file names, sizes and accessibility findings — for 30 days. We keep no copies of your documents or your pages.
• We back off. A 429 or 503 drops us to one request at a time; a bot challenge stops the scan rather than retrying against it.

Where it comes from

Requests originate from Cloudflare Workers, so they arrive from Cloudflare's shared edge rather than a fixed range we could publish. Identity is proved cryptographically instead: requests carry an HTTP Message Signature (RFC 9421) and our public keys are published at/.well-known/http-message-signatures-directoryon audit.livingpage.app.