Skip to article
auraGEO
IN PRACTICE · Your website and AI search

Your website is open to people but closed to crawlers: how to check CDN and WAF rules

Compare the page response with security events in the CDN. A permissive robots.txt does not help if a WAF, browser challenge or rate limit blocks the request earlier. Fix the specific rule and access scenario.

Why the owner’s browser is not a test

The owner may be signed in, have an existing cookie or be on an allowed network. A crawler arrives without these conditions. A page that opens for you can therefore return 403, 429 or HTML asking another client to prove it is human. Inspect the response content, not just the final status code.

Match the failed retrieval time against the security log: which path, action, rule and category triggered? Do not conclude “the provider is blocked” from a single request from your own server. That is an observation about one test location and method.

Review the current AI bot policy

Cloudflare documents separate Search, Agent and Training policies. Settings and mixed-purpose crawler behavior depend on the selected configuration; an old name for a general toggle does not explain all access. In addition to AI settings, custom WAF rules, rate limits and protection for particular paths may apply.

For an exception, use criteria supported by the platform and a precise scope. Allowing requests based only on an arbitrary User-Agent string is easy to imitate. Verified client information and the provider’s official IP ranges are more reliable. Agree on the change with the person responsible for security and load.

Verification after a change

First save the original denial, then make one targeted edit and repeat the same request. Open the security events again: the request should pass under the expected rule and return the main content. Make sure the workspace and administrative URLs have not been opened.

For example, an exception for reading public documentation does not need to affect checkout. This narrow approach lets you assess the outcome and quickly return to the previous configuration if load increases unexpectedly.

AURA’s role in diagnosis

AURA GEO’s technical audit saves responses from accessible URLs and synthetic probes. The agent uses these for the task, while CDN logs and actual server events help pinpoint the cause.

MCP does not grant permission to change someone else’s firewall. The owner provides settings access separately. Once the block is removed, continue checking discovery and citation: a properly configured WAF solves access, not the model’s choice of source.

Conditions and limitations
  • A synthetic probe does not reproduce the provider’s network.
  • Open access does not mean indexing.
PRIMARY SOURCES

Explore the details

  1. Cloudflare: AI bot access policies
  2. Cloudflare: targeted rule exceptions
NEXT STEP

The reason for the denial is confirmed and a narrow fix is verified.

Prepare an audit for your agent
All articles