Only 12% of the world's top websites are ready for AI agents

AI agents — ChatGPT, Claude, Perplexity, and the growing class of autonomous MCP clients — don't browse the web like humans. They look for specific, machine-readable signals: llms.txt, agents.txt, an x402 payment manifest, and a handful of others. If your site lacks them, agents literally cannot discover or use it.

So we asked a simple question: are the world's biggest websites ready? We ran a deterministic audit of 143 major domains across 8 agent-discovery surfaces. Here's what we found.

The surprise: it's not about robots — it's about llms.txt

Almost every major site still has a robots.txt (93%) and most publish security.txt (85%). But the one file specifically designed to tell LLMs what your site is about — llms.txt — is present on just 12% of top sites. That's the gap.

Meanwhile, agent-payment and agent-card standards have quietly gone mainstream: roughly 72% of these sites already expose an x402 manifest or an agents.json/agents.txt. The infrastructure for agents to pay and coordinate is further along than the infrastructure for them to understand.

What this means for your site

If you run a website, the takeaway is practical:

• Add llms.txt — it's the single biggest differentiator right now, and almost nobody has it.
• Add agents.txt and .well-known/agents.json — describe what agents can do with your site.
• Consider x402 if you want AI agents to be able to pay you directly.

See the full ranking

The complete, sortable leaderboard of all 143 sites — with per-site grades and which surfaces each one passes — is live:

→ Open the AI Visibility Index

We re-audit continuously. If you'd rather have an expert do it for you, the weekly AI Visibility Report (€19/mo) tracks your own site over time.

Methodology: a surface passes if it returns HTTP 200–399 with a non-empty body. Snapshot of 143 major domains. Deterministic, no AI used in scoring.