See what AI sees — on the record.

Hidden Layer is an independent publication and audit tool that measures how AI systems see — and miss — public websites. Output is a single 100-point agent-readiness score across 46 checks and eight categories. Free to use, reproducible from public infrastructure, no account required.

Hidden Layer audits the gap between what a website publishes and what AI crawlers, language models, and shopping agents actually do with it. Every audit runs the same 46 checks across eight categories — discoverability, bot access, AI discovery, AI visibility, agent integration, commerce, product pages, and GEO presence — and returns a single 100-point score on an A–F scale.

The publication treats the audit corpus as its source material. Posts cite their own findings; methodology is open; the stack is free-tier only. Three publishing slots a week, on the same domains anyone else can audit. The product is the credential.

Hidden Layer began as a question: when a brand says they are doing well in AI search, how would they actually know? The agent-readiness audit is the answer. Forty-six checks across eight weighted categories, applied identically to every domain, normalised to a 100-point scale, and surfaced with the underlying evidence. No black box, no proprietary feed, no embargoed dataset behind a paywall.

The categories cover what AI systems actually use: robots.txt and sitemap discoverability, the 12 canonical AI bot User-Agents, llms.txt and llms-full.txt publication, JSON-LD structured data quality, entity linking via sameAs and Wikidata, well-known agent files, x402 payment signaling, Google Merchant Center feed readiness, product schema completeness, and cold-recall presence in LLM training data via Common Crawl and Wayback Machine signals.

The publication ships roughly three posts a week. Each one is grounded in the same audit pipeline that powers the free tool — never speculation, always reproducible. The methodology lives in the audits, not in the bios. Anyone can run the same checks on their own domain, get the same score, and read the breakdown. That is the whole proposition.

Interview, quote, and comment requests are handled via the publication desk rather than individual contributors. Response within two business days.

• hidden-layer-logo.svgpending upload
• hidden-layer-wordmark.svgpending upload
• hidden-layer-press-kit.pdfpending upload