Skip to content
INTERNET INFRASTRUCTURE & NETWORKS

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

The rapid proliferation of AI-driven personal assistants is fundamentally shifting how users interact with the internet. Acting as intermediaries between humans and the digital landscape, these personal agents translate human instructions into direct requests and actions across websites. Whether a customer asks an AI assistant to compare products, locate support articles, or seamlessly complete a purchase, the nature of website traffic is evolving. For site owners, however, these interactions often appear simply as automated traffic. Without clear visibility into which agents are visiting and what tasks they are executing, determining whether this traffic benefits a business or introduces security risks becomes a major challenge.

While blocking personal agents entirely may seem like the safest preventative measure to protect digital assets, it often results in alienating potential customers who rely on these tools. Recognizing this delicate balance, Cloudflare has announced the launch of new analytics capabilities designed to help webmasters understand how personal agents access their sites on behalf of humans. Alongside these analytics, the company is introducing more granular access policies to put those insights into practice, empowering site owners to welcome AI agents on their own terms and in ways that actively support their underlying business models.

See how agents access your site

Understanding which specific agent is behind an automated request transforms ambiguous traffic into actionable data that security and business teams can manage with confidence. Cloudflare maintains one of the largest verified bot directories in the world, and the company is now integrating that intelligence directly into everyday traffic monitoring tools. Within Security Analytics, the Bot Analysis chart now features a dedicated category for Personal Assistants, sitting alongside existing classifications such as Verified Bot, Likely Human, Likely Automated, and Automated. Security teams can isolate this category to track personal agent traffic over time with precision.

Further expanding on this visibility, the Top statistics interface now includes a specialized Personal Assistants widget. This component lists every individual agent visiting a site and details the precise volume of requests each one sends. When an administrator filters by a specific agent, all other associated widgets update dynamically to reflect only that agent’s traffic details. This includes the paths requested, the HTTP status codes received, and the originating networks. At a glance, administrators can discern whether an agent is successfully logging in, browsing product pages, navigating to checkout, or encountering errors on pages it lacks permission to access.

For websites that utilize Account Abuse Protection—a feature currently in early access—site owners and fraud investigation teams gain deeper insight into precisely which user account each personal agent represents. The population overview page illustrates which personal agents are active across a site’s user base and quantifies how many accounts rely on them, enabling businesses to evaluate whether AI agent traffic constitutes a meaningful commercial segment. Every login and sign-up event now explicitly indicates whether a personal agent initiated the request and identifies which one. Fraud teams can filter accounts based on AI assistant events or the sheer number of associated assistants.

Furthermore, individual user account profiles display the specific agents operating on behalf of that account. When fraud analysts investigate unusual login behavior, they can immediately determine whether a request originated from an agent the customer has utilized for weeks or from an unfamiliar agent that appeared for the first time on that day. If an attacker appears to be weaponizing an agent against a specific account, fraud analysts can apply targeted mitigations, such as implementing a Web Application Firewall rule tied to the Hashed User ID, rather than executing a broad block that would penalize legitimate users relying on the same agent. Together, these comprehensive views provide a robust foundation for deciding which personal agents to welcome, restrict, or block.

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

Set access policies that reflect your preferences

Earlier in the year, Cloudflare introduced dedicated AI bot policies to allow site owners to establish distinct access control rules for AI bots utilized across Search, Training, and Agent behaviors. Building upon that framework, the company is expanding the existing Agents policy to incorporate granular control options specifically tailored for supported personal agents. To accurately reflect this broader operational scope, the setting has been officially renamed to AI Crawler and Agent access policies.

Within the Security Settings dashboard, administrators can configure an overarching Agents policy and subsequently define tailored rules for each supported personal agent. Available configuration choices include allowing access by choosing not to block, blocking access exclusively on pages that feature advertisements, or implementing a complete block. For instance, a site owner might configure the general Agents policy to block traffic on ad-supported pages while simultaneously selecting an allow option for a specific, trusted personal agent. Existing Search and Training policies remain unaffected by these updates.

These access preferences can also be configured during the initial onboarding process when adding a new zone to Cloudflare. During setup, administrators are asked whether their pages are monetized with advertisements, allowing the system to tailor recommended policies accordingly. These controls can be modified at any time within the Security Settings interface. Ultimately, the appropriate policy depends heavily on the unique business model of each website. One publisher may choose to permit a supported personal agent across the entire domain to help readers discover articles, while another publisher may restrict that same agent on pages where advertising revenue sustains the content. Both approaches allow businesses to enforce access controls aligned with their commercial objectives.

Agent identity is only part of the story

While the new analytics and policies provide site owners with unprecedented visibility and control over personal agent activity, identifying the agent is only part of the broader challenge. Cloudflare’s next strategic objective is to help webmasters preserve the vital context of their relationships with users even when personal agents act as intermediaries. Frequently, a website may successfully recognize the agent making a request while lacking crucial contextual information about the underlying user, such as whether they are an existing customer, what specific actions they authorized the agent to perform, whether they maintain an active subscription, or which security protections should apply.

Achieving this level of context preservation will require privacy-aware mechanisms capable of carrying user context securely through an agent, helping websites accurately assess reputation, and establishing clear accountability when operational errors occur. Cloudflare plans to share further developments as it collaborates with agent operators and industry participants across the ecosystem. This ongoing collaboration aims to ensure site owners can confidently grant agents greater access while continuing to foster direct relationships with their human users.

Make the choice for your site

Ultimately, a person asking an AI assistant to compare products, locate support documentation, or complete a transaction remains a customer trying to engage with a business. Website operators are encouraged to review how personal agents are currently accessing their infrastructure and visit Security Settings within the AI Crawler and Agent access policies dashboard to establish terms that best serve their operational needs.