The State of the Agent-Readable Web

// Missing llms.txt

63%

63% of the top 390 AI companies have no llms.txt or one too thin for inference. This is the single cheapest agent-readability fix — and most of the industry still skips it.

63% of the top 390 AI companies are missing or under-serving llms.txt.
Source: grow.contact/stats#missing-llms-txt (CC BY 4.0)

// Weak JSON-LD

67%

67% ship insufficient structured data for reliable AI citation. ChatGPT, Perplexity, and Google AI Overviews lean on JSON-LD to know who, what, and how much.

67% of top AI companies ship insufficient JSON-LD for reliable AI citation.
Source: grow.contact/stats#weak-jsonld (CC BY 4.0)

// Opaque to AI engines

27%

27% of the sample score below 55/100 — effectively invisible to ChatGPT, Perplexity, and Claude live search. Most are fixable in a single sprint.

27% of top AI companies are effectively opaque (score <55/100) to ChatGPT, Perplexity, and Claude.
Source: grow.contact/stats#opaque (CC BY 4.0)

// Clear the agent-native bar

13%

Only 13% of the sample clear 85/100 — the bar where a site is reliably cited across all four major AI search engines.

Only 13% of top AI companies clear the agent-native bar (score ≥85/100).
Source: grow.contact/stats#agent-native (CC BY 4.0)

// Fail the speed threshold

66%

66% of sites are slow enough that AI crawlers (which timeout in 1–5s) skip them outright. The penalty isn't a ranking drop — it's silence.

66% of top AI companies fail the first-byte speed threshold AI crawlers time out against.
Source: grow.contact/stats#slow-ttfb (CC BY 4.0)

// Weak semantic HTML

69%

69% miss core landmark elements (<main>, <article>, <nav>) AI scrapers use to extract content reliably. Cheap to fix; rarely audited.

69% of top AI companies ship HTML without the semantic landmarks AI scrapers rely on.
Source: grow.contact/stats#weak-semantic (CC BY 4.0)