Skip to the yard
Waxsealai

Cloudflare Adds Personal Agent Analytics and Per-Agent Access Controls

Shipped
Shipper
Marina Elmore
Readers
797

Cloudflare’s new dashboards identify personal agent traffic, while per-agent controls let site owners allow or block access based on their needs.

Cloudflare Adds Personal Agent Analytics and Per-Agent Access Controls

Personal agents can browse product pages, find support information, and make purchases for people. For a website, those actions arrive as automated requests, which can be hard to distinguish from other bot traffic. Cloudflare is adding analytics to identify personal agents and controls to set access rules for individual supported agents. The change gives site owners more options than treating all agent requests alike.

The practical value depends on visibility. If a site owner can see which agent sent a request, which pages it reached, and what responses it received, they can make access decisions with more context. A stream of requests might represent customers researching products, or agents repeatedly reaching pages they cannot use. The new tools are intended to help teams tell those cases apart before choosing whether to allow, limit, or block access.

Agent activity appears in traffic analytics

Cloudflare is adding a Personal Assistants category to the Bot Analysis chart in Security Analytics. It sits alongside Verified Bot, Likely Human, Likely Automated, and Automated. Selecting the new category filters the chart to personal agent traffic over time. The classification draws on Cloudflare’s verified bot directory.

A new Personal Assistants widget in Top statistics lists the agents visiting a site and the number of requests each sends. Selecting an agent filters the other widgets to show details for that traffic, including requested paths, response status codes, and originating networks. That makes it possible to check whether an agent is reaching login pages, product information, checkout, or pages that return errors.

Those details can guide troubleshooting as well as access decisions. For instance, repeated errors on a particular page may indicate that an agent is trying to use a part of the site that does not support its requests. The analytics show where the interaction is occurring and what response the site is returning; site owners can then decide whether the traffic fits their goals.

Account tools add user-level context

For sites using Account Abuse Protection, which is in early access, Cloudflare says site owners and fraud teams can see which user an agent represents. A population overview shows the personal agents active across a site’s users and how many accounts use them. Login and sign-up events also identify whether an agent made the request and name the agent. Teams can filter accounts by AI assistant events or by the number of assistants associated with them.

Each user’s account overview lists the agents acting for that account. This can help an analyst compare a suspicious login with the account’s past activity: an agent used regularly by a customer provides different context from one that appears for the first time. If an account appears to be under attack, analysts can apply a Web Application Firewall rule using the account’s Hashed User ID. That lets them act on the account without blocking the agent for every other customer.

The distinction matters because an agent’s identity alone doesn’t establish whether a request is appropriate. Account history can help teams assess activity in context, while account-level controls offer a way to respond to a specific concern without imposing the same restriction across all users.

Policies can vary by supported agent

Cloudflare previously introduced separate access choices for AI bots used for Search, Training, and Agent behaviors. The existing Agents setting is now expanding to provide more granular controls for supported personal agents, under the name AI Crawler and Agent access policies. The Search and Training settings remain unchanged.

In Security Settings, site owners can select an overall Agents policy and then set a different choice for each supported personal agent. The options are Allow (do not block), Block on pages with ads, and Block. For example, a site could block agents on pages with ads by default while allowing a particular supported agent across the site.

The same decisions are available when adding a new zone to Cloudflare. During onboarding, Cloudflare asks whether the site earns revenue from ads and uses the answer to tailor recommended settings. Site owners can review or change those settings later in Security Settings.

These controls let different sites make different choices about the same agent. A publisher may allow access so readers can find its work through an assistant, while another may restrict access on pages where advertising supports the content. The available settings provide a way to express that distinction, though the right configuration depends on how each site serves its users and funds its pages.

Identity does not carry the full user relationship

Cloudflare describes agent identification and access control as an initial step. A site may recognize the software making a request but still lack information it would normally use to understand the person behind it. That can include whether the person is already a customer, what they authorized the agent to do, whether they have a subscription, and which protections should apply.

Addressing that gap will require privacy-aware ways to pass user context through agents, assess reputation, and establish accountability when problems occur. Those pieces are not included in the controls described here. Until sites can evaluate more of the user’s context, the new analytics can clarify what agents are doing, while access settings let site owners define the boundaries they’re comfortable with.

Site owners can review agent activity and update their choices under Security Settings → AI Crawler and Agent access policies. The immediate takeaway is practical: measure which agents reach the site, examine their request patterns, then set access by agent where Cloudflare supports it.

Source: Marina Elmore · blog.cloudflare.com

Same vesselRelated cargo