This site runs on what it sells

An AI crawler that's blocked in robots.txt won't index or cite the page at all, regardless of how good the content is. This is the single most binary check in the report.

If a site serves meaningfully less content to a known AI crawler than to a regular browser, that crawler can't see (and can't cite) what a human visitor sees. This can happen by accident — a bot-detection or security tool treating unfamiliar user agents as suspicious.

Many AI crawlers fetch raw HTML and never execute JavaScript. If the meaningful content only appears after client-side rendering, those crawlers see an empty shell — this is the single most common real failure in AI visibility.

AI crawlers and answer engines use heading structure to understand how a page is organized and what each section is about. Multiple or missing <h1>s and skipped levels make that structure ambiguous.

AI answer engines favor content that states the answer plainly before elaborating. A page that opens with navigation, a hero banner, or throat-clearing before any substantive text is less likely to be quoted directly.

JSON-LD tells an AI engine explicitly what a page is — an article, a product, an FAQ — rather than making it infer that from prose. It's the clearest signal a page can send about its own content.

These tags are the fastest way for any crawler — search or AI — to understand what a page is about without parsing the full body. Missing or duplicate tags create ambiguity about which URL is authoritative.

A sitemap gives crawlers a direct list of pages to check, rather than relying entirely on discovering links by crawling. It's low-cost and widely supported.

llms.txt is a low-cost, forward-looking signal, not a guarantee of anything. No major AI engine has confirmed it drives citations, and Google's own guidance is that site owners don't need special files for AI Search. It's worth having as part of a bundle — it costs little — but the load-bearing work is crawler access, server-rendered content, and structured data, not this file.

AI agents (and screen readers) don't see pixels — they read the accessibility tree: element roles, names, and states. An unlabeled icon button or an input with no associated label is invisible to them even though a sighted human can infer its purpose. This is exactly what Google's Lighthouse Agentic Browsing category audits, and for a product about AI readability specifically, failing it undermines the pitch.

Slow responses cost crawl budget — crawlers (AI or otherwise) fetch fewer pages per visit when each one is slow, and some abandon slow requests entirely.