agentspeed.
← all tools
Tool

AI Crawler Access Matrix

Fetch any URL in parallel as ten AI crawlers and a default browser. See exactly which agents your site allows, blocks, redirects, challenges, or serves a different body to. Real HTTP fetches, real verdicts.

Tool
›
https://amazon.com10 crawlers tested1 allowed2 different7 blocked
agentverdictstatusbyteslatencysimilaritynote
GPTBotblocked5032,671263ms0.00OpenAI training crawler
ChatGPT-Userblocked503559291ms0.00ChatGPT browsing for a user
ClaudeBotblocked5032,671261ms0.00Anthropic crawler
Claude-Userblocked5032,671244ms0.00Claude browsing for a user
PerplexityBotblocked5032,671247ms0.00Perplexity citation crawler
Google-Extendedblocked5032,671249ms0.00Google AI training opt-out crawler
Googlebotallowed · diff2001,386,506278ms0.11Google search index crawler
Bingbotallowed · diff200650,929518ms0.76Bing search index crawler
CCBotblocked5032,671290ms0.00Common Crawl, feeds many AI training datasets
Browserallowed200796,784796ms1.00Vanilla desktop Chrome, control row

What each row means

  • GPTBotblockedHTTP 503. The host is refusing this user-agent.
  • ChatGPT-UserblockedHTTP 503. The host is refusing this user-agent.
  • ClaudeBotblockedHTTP 503. The host is refusing this user-agent.
  • Claude-UserblockedHTTP 503. The host is refusing this user-agent.
  • PerplexityBotblockedHTTP 503. The host is refusing this user-agent.
  • Google-ExtendedblockedHTTP 503. The host is refusing this user-agent.
  • Googlebotallowed · diffAllowed (2xx) but the body differs substantially from the browser baseline. Possible cloaking, agent-aware rendering, or paywall variant.
  • Bingbotallowed · diffAllowed (2xx) but the body differs substantially from the browser baseline. Possible cloaking, agent-aware rendering, or paywall variant.
  • CCBotblockedHTTP 503. The host is refusing this user-agent.
  • BrowserallowedBaseline browser response, used as the comparison anchor.
Result summary

Copyable summary for sharing with infra owners or filing an issue with your CDN.

AgentSpeed AI Crawler Access Matrix
Target: https://amazon.com
Run: 2026-09-04T04:10:23.602Z

Per-agent verdict:
  - GPTBot           blocked        status=503 sim=0.00 bytes=2671
  - ChatGPT-User     blocked        status=503 sim=0.00 bytes=559
  - ClaudeBot        blocked        status=503 sim=0.00 bytes=2671
  - Claude-User      blocked        status=503 sim=0.00 bytes=2671
  - PerplexityBot    blocked        status=503 sim=0.00 bytes=2671
  - Google-Extended  blocked        status=503 sim=0.00 bytes=2671
  - Googlebot        allowed · diff status=200 sim=0.11 bytes=1386506
  - Bingbot          allowed · diff status=200 sim=0.76 bytes=650929
  - CCBot            blocked        status=503 sim=0.00 bytes=2671
  - Browser          allowed        status=200 sim=1.00 bytes=796784

Run again at: https://agentspeed.com/tools/ai-crawler-access-matrix
Watch this URL to get alerted when this changes

Turn this one-off check into a scheduled watch. We re-run it, diff the result, and alert you on a regression through the same channels as your agent alerts. Free includes one weekly watch; paid plans add more and choose their own schedule.

Manage watches at /watches. You’ll be asked to sign in if you haven’t already.

Examples

Click an example to run the tool with pre-filled inputs.

Why this matters for AI agents

Most agent-readiness problems show up here first. A site that blocks GPTBot at the WAF, redirects ChatGPT-User to /login, or serves a different body to PerplexityBot than to a browser will never show up in agent answers, and the only way to find out is to actually fetch as those agents.

This is the live equivalent of the journey.primary_probe check on a public scan, run on demand against any URL.

Common fixes
WAF blocks GPTBot
In Cloudflare bot management, allow GPTBot, ClaudeBot, and PerplexityBot explicitly. Their UA strings are documented and IPs are fixed.
Different body for bots than browser
If similarity drops below 0.80 between an agent and the Browser baseline, your CMS or edge layer is likely rendering a different variant. Audit any UA-conditional rendering rules, since agents often get the worst one.
Redirected to /login
Render server-side preview content for unauthenticated requests on landing pages. Agents cannot authenticate; they cite the auth wall.
Cloudflare challenge fires
Add the bot to Cloudflare’s allowed-bots list, or whitelist the published IP ranges for OpenAI, Anthropic, Perplexity.
Different canonical per UA
Canonical drift confuses agent dedupe. Audit any UA-conditional canonical rules; the canonical should be identical across all UAs.
Linked rubric checks
FAQ
Is this the same as the Agent UA tester?
Same backend pipeline, broader UA list. The UA tester runs 6 UAs to focus on cloaking detection. The Crawler Access Matrix runs 10, adding Claude-User, Google-Extended, Googlebot, Bingbot, and Common Crawl, to give a full picture of who can and cannot reach your site.
Why include Googlebot and Bingbot?
Because misconfigured WAFs frequently block them too. If your search visibility tanks, the matrix tells you whether the cause is your robots.txt, your bot manager, or your CDN, without you having to spelunk through logs.
What does "allowed-different" mean?
The agent got a 200 OK but the body content differs from the browser baseline by more than 20% (Jaccard similarity < 0.80). This is usually cloaking, paywall variants, or "render lite for bots" rules. Agents see worse content than humans.
Do you store the responses?
No. Each run is fetched live and discarded. The result page is rendered server-side; nothing persists.
How do redirects show up?
fetchWithHostPolicy follows up to 5 hops. The "redirected" column shows whether any redirect happened; the final URL column shows where it ended up. SSRF-protected, so every redirect is re-validated against the host policy.
Run a full agent-readiness scan
Full scan

Score the entire site against 37 agent-readiness checks across discoverability, readability, structured data, actionability, and performance.

Get the full report for amazon.com →