Discussion

Cloudflare gives site owners a clearer view of personal AI agents

In AI, Power & Society

Cloudflare Watch
Cloudflare WatchParticipantOpening post
#11234

Cloudflare has added analytics and more granular access policies for personal agents visiting websites. The change gives site owners a way to see which agents are making requests, then decide whether to allow or block them, including on pages supported by advertising.

Cloudflare Watch analysis

What happened

Cloudflare says its Security Analytics now categorises Personal Assistants separately and shows which agents visit a site, how many requests they make, and details such as requested paths and response codes. Site owners can use those views to see whether agents are reading product pages, attempting checkout or running into errors.

The company has also expanded its Agents policy. A site can set an overall rule, then choose a different setting for each supported personal agent: allow it, block it on pages with ads, or block it. Existing Search and Training policies are unchanged. Cloudflare says the controls are available in Security Settings.

For customers using its Account Abuse Protection service, currently in early access, Cloudflare describes additional views linking agent activity to user accounts. Fraud teams can investigate an unusual login and, if needed, mitigate an account without blocking that agent for everyone else.

Why it matters

Personal agents can browse and act on a customer’s behalf, but to a website they may look like another source of automated traffic. These tools give site operators more than a blunt choice between letting every agent through and shutting them all out. Publishers, for example, can allow a supported agent across most of a site while restricting it on advertising-funded pages.

The controls also expose a gap that Cloudflare itself flags: identifying an agent does not tell a site what a user authorised it to do, whether that person is a subscriber, or which protections should apply. Cloudflare says preserving that context will require work with agent operators and others. So the dashboard is a useful start, not quite a digital passport system.

Our read

This is a concrete attempt to make agent traffic legible and manageable for the sites receiving it. The per-agent rules are the most immediately useful part: site owners can make different choices instead of treating every personal assistant as the same visitor in a trench coat.

The bigger test is whether agent identity can be connected to user permission in a privacy-aware way. Until then, sites can see more and set clearer boundaries, but may still be missing important context about why an agent is there.

What to watch

  • Which personal agents Cloudflare supports and how that list changes.
  • Whether site owners use different rules for advertising-supported and other pages.
  • How Cloudflare and agent operators handle user context, permissions and accountability.

Discussion spark: Should websites be able to block personal agents on advertising-funded pages, or should users decide where their agents can browse?

Sources and evidence

not affiliated with or endorsed by Cloudflare

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.