Skip to content
INTERNET INFRASTRUCTURE & NETWORKS

Cloudflare Introduces Advanced Analytics and Granular Access Controls to Manage Personal AI Agent Traffic

As artificial intelligence transforms how humans interact with the web, website owners face a growing operational challenge: distinguishing between beneficial automated traffic and potential security risks. Personal agents—software intermediaries that translate human instructions into requests and actions on websites—are becoming increasingly common. Whether a customer asks an AI assistant to compare products across an e-commerce catalog, locate specific support documentation, or execute a purchase, these interactions arrive at web servers as automated traffic.

For site operators, managing this traffic has historically been a blunt-force dilemma. Without clear visibility into which specific agents are visiting and what tasks they are performing, it is difficult to determine whether automated interactions are driving valuable business or introducing security vulnerabilities. While blocking all automated agents can seem like the safest preventative measure, it often comes at the steep cost of alienating legitimate customers who rely on these tools.

To address this shifting digital landscape, Cloudflare has announced a suite of new analytics and granular access policies designed to help website owners understand and manage how personal agents access their web properties. By providing deep visibility into agent behavior and offering flexible control mechanisms, the new features allow organizations to welcome automated assistants on their own terms and in ways that directly support their commercial objectives.

See how agents access your site

Understanding the nature of incoming web traffic is the first step toward effective management. When automated requests hit a website, knowing which agent is behind the activity turns an unexplained surge of traffic into actionable data. Leveraging its extensive verified bot directory—the largest of its kind in the world—Cloudflare is now integrating this intelligence directly into individual traffic analytics dashboards.

Within the Security Analytics interface, the Bot Analysis chart now features "Personal Assistants" as a dedicated category, appearing alongside established classifications such as Verified Bot, Likely Human, Likely Automated, and Automated. Security teams can filter their views specifically for personal agent traffic, observing trends and patterns over time with unprecedented clarity.

Beyond high-level trends, the platform introduces a new Personal Assistants widget within the Top statistics section. This tool lists every specific agent visiting a site alongside the exact volume of requests it generates. When an administrator filters by a single agent, the entire dashboard dynamically updates to reflect only that specific entity’s traffic details. Site operators can instantly review the request paths being targeted, the HTTP status codes returned, and the originating networks. At a single glance, teams can verify whether an agent is successfully logging in, reading product documentation, navigating toward the checkout pipeline, or encountering errors on sections of the site it is not authorized to access.

For organizations utilizing Account Abuse Protection—a security suite currently in early access—these analytics extend down to individual user accounts. Website operators and dedicated fraud teams gain the ability to map which personal agents are actively representing their user base and determine how many distinct accounts utilize them. Every login and sign-up event is now tagged to show whether a personal agent initiated the request and which specific tool was employed. Fraud analysts can easily filter accounts by AI assistant activity or evaluate the total number of connected assistants per user profile.

Furthermore, individual account overviews display the specific agents operating on behalf of a given user. If a fraud analyst investigates an anomalous login event, they can immediately discern whether the request originated from a trusted agent the customer has relied upon for weeks or from an unfamiliar assistant that appeared for the very first time. If an attacker attempts to exploit an account by weaponizing an agent, fraud analysts can apply targeted mitigations—such as a Web Application Firewall (WAF) rule targeting the specific Hashed User ID—without resorting to a blunt platform-wide block that would disrupt legitimate users relying on the same assistant.

Set access policies that reflect your preferences

Visibility alone is only part of the equation; website operators also require the tools to enforce their business preferences. Earlier this year, Cloudflare introduced specialized AI bot policies to give site owners separate access control choices for AI crawlers used for search engine indexing, model training, and automated agent behaviors.

See the agents, set the terms: new visibility and control for personal agents

To reflect a broader scope of functionality, Cloudflare is expanding its existing agents policy to encompass granular controls for supported personal agents, renaming the overall configuration setting to "AI Crawler and Agent access policies." Found within Security Settings, this feature allows administrators to establish a baseline policy for general agents while defining custom rules for specific, supported personal assistants.

Site owners can choose from three distinct policy options for individual agents: allowing access without restriction, blocking access exclusively on pages containing advertisements, or implementing a complete block. For instance, an organization might configure a blanket policy that blocks generic agents from ad-supported content while explicitly permitting a favored personal assistant to traverse the site freely. Existing policies concerning search engine crawlers and data training bots remain unaffected by these updates.

These preferences can also be established during the initial onboarding process when a new domain is added to Cloudflare. By asking site owners whether they monetize their pages through advertising, the platform tailors its initial policy recommendations accordingly, though administrators retain the freedom to modify their access controls at any time through the Security Settings dashboard.

This flexibility acknowledges that a one-size-fits-all approach is inadequate for the modern web. One digital publisher may choose to permit a supported personal agent across all pages to ensure readers can seamlessly discover and consume articles, while another publisher might restrict that same agent on revenue-generating, ad-supported pages. Both approaches allow businesses to align technical security controls with their unique operational models.

Agent identity is only part of the story

While the new analytics and access policies provide significant improvements in visibility and control, industry observers note that managing agent identity represents only an initial phase in a much larger transition. As personal agents increasingly act as intermediaries between human consumers and digital services, preserving the context of the relationship between the user and the website remains a critical hurdle.

Currently, a website may successfully recognize the automated agent making an incoming request, but it often lacks the nuanced context it would normally possess during a direct human interaction. Web servers may not immediately know whether the agent represents an established, paying customer, what specific permissions the user granted to the assistant, whether an active subscription is in place, or which specific compliance protections should apply to the session.

Bridging this gap will require the development of privacy-aware mechanisms capable of securely carrying user context through an intermediary agent, helping websites evaluate agent and user reputation objectively, and establishing clear lines of accountability when transactions encounter errors. Cloudflare indicates it will continue collaborating with agent operators and other stakeholders across the digital ecosystem to address these challenges, ensuring site owners can make informed decisions about granting broader access while nurturing direct relationships with their human audiences.

Ultimately, the underlying intent remains unchanged: the person utilizing a personal assistant to compare consumer products, track down customer support articles, or complete a financial transaction is still a customer trying to engage with a business. By reviewing how personal agents access web properties and configuring targeted policies within the Security Settings interface, businesses can establish operational terms that protect their digital assets while remaining open to the future of consumer technology.