What do AI agents see on your site?

One URL, two fetches, full Agent Experience diff.

Agent Experience (AX) is the developer-facing companion to UX — the experience an AI agent has when it visits your site. Modern sites quietly serve different responses to agents than to humans (smaller markdown payloads, stripped HTML, or sometimes nothing at all). This tool requests the same page twice (once as the agent, once as a vanilla browser) and shows you what came back, side by side.

1. 01

   Pick a URL and an agent

   Choose Claude Code, GPTBot, PerplexityBot, ChatGPT-User, or paste a custom User-Agent. Optionally send Accept: text/markdown the way a real agent would.

2. 02

   Two parallel fetches

   A Cloudflare Worker hits the URL twice at the same time — once with the agent's identity, once as a vanilla browser.

3. 03

   Diff body, headers, and signals

   Status, bytes, time, body similarity. Plus the off-page hints: /llms.txt, /llms-full.txt, robots.txt per-bot policy, Vary: User-Agent, Content-Signal, and rel="alternate" markdown links.

4. 04

   Read the verdict

   A one-line summary tells you whether the agent got a smaller markdown payload, the same HTML a human sees, a redirect, or a block — with the response panes underneath so you can scan the actual bytes.

Around 97% of sites still ship no agent-specific signals, and HTML costs roughly 3× the tokens of equivalent markdown — so most improvements are quick wins. Start here:

• Cloudflare “Markdown for Agents” — one toggle if your site is on Cloudflare; serves markdown on Accept: text/markdown.
• llms.txt specification — an emerging convention for pointing agents at your most important content. Adoption is still patchy, but it’s a cheap signal.
• Bot allowlist (robots.txt) — the precondition. If ClaudeBot or GPTBot is disallowed, nothing else matters.
• agentexperience.ax — the AX manifesto and principles, by the team at Netlify.