On January 23, 2025, OpenAI introduced Operator, a groundbreaking AI system that can control a web browser to perform tasks on behalf of users. This marks a significant evolution in how AI interacts with the web—moving from passive crawling to active browsing.

While this advancement brings exciting possibilities, it also presents a fundamental challenge: How can websites understand and serve the actual user behind the AI agent?

The Challenge of User Context

When Operator visits a website, it appears to come from Microsoft Azure's IP addresses, as it runs in OpenAI's cloud infrastructure. This creates a significant disconnect for websites trying to serve their users effectively. Instead of seeing the actual user's location, websites see Azure datacenter IPs. This affects:

  • Content personalization
  • Security measures
  • Geographic services defaulting to datacenter locations
  • Rate limiting systems misfiring

This challenge isn't new—it’s similar to issues websites face with CDNs and proxy services. Fortunately, there's a well-established solution: The X-Forwarded-For (XFF) header.

Understanding X-Forwarded-For

XFF has a rich history in web infrastructure. Originally introduced by the Squid proxy server in the 1990s, it was designed to track original client IPs through proxy chains. Over time, it evolved into a widely adopted standard used by CDNs, load balancers, and proxy services worldwide.

A typical XFF header looks like this:

X-Forwarded-For: original_client_ip, proxy1_ip, proxy2_ip

For Operator, OpenAI simply adds the user’s real IP to outbound requests:

X-Forwarded-For: user_real_ip

This allows websites to identify the actual user making the request, even though it's coming through OpenAI's infrastructure.

Current Website Behavior

Websites handle XFF headers in different ways:

  • Ignore XFF Completely: Only sees Azure IPs, loses user context.
  • Basic XFF Processing: Extracts first IP, uses for geolocation and basic rate limiting.
  • Sophisticated Implementation: Validates proxy chain, preserves full user context, enhances security.

Why It Matters for AI Agents

OpenAI's inclusion of XFF headers in Operator sets a precedent for AI agent transparency. This allows websites to:

  • Maintain security measures
  • Provide personalized, location-aware content
  • Ensure accurate analytics even when requests come through an AI intermediary

Looking Forward

As more AI agents emerge, we expect:

  • Industry Standardization: More agents adopting XFF headers and standardized proxy chain formats.
  • Enhanced Services: Better geographic targeting, improved personalization, more accurate analytics.

A New Era of Analytics with Profound

At Profound, we've built a comprehensive platform for tracking and analyzing AI agent traffic. Our platform provides:

  • Real-time AI agent traffic monitoring
  • Original user location tracking
  • Behavioral pattern analysis
  • Security and rate limiting insights
  • Performance metrics and optimization recommendations

today with Agent Analytics to learn more about how your website interacts with AI agents.