4 November 2026
Cloudflare's launch post describes a Personal Assistants category and per-agent traffic views in Security Analytics. The practical question for a site owner is which requests to permit, and how to verify that the resulting policy works.
What changes for site owners
Under Security → Settings → AI Crawler and Agent access policies, supported personal agents can receive individual actions: allow, block on pages where Cloudflare detects ads, or block. Cloudflare's policy documentation says personal agents initially inherit the overall Agents policy. Changing that overall policy updates the personal-agent settings, so review individual choices again afterward.
The launch also describes account-linked agent views through Account Abuse Protection. That product remains Early Access for Bot Management Enterprise customers, with access requested through the account team. User ID requires opt-in. Its availability documentation describes no additional charge for a limited early-access period. It should not be read as a permanent free feature for every plan.
Analytics depth also varies. Cloudflare's Bot Analytics documentation distinguishes Business and Enterprise customers without Bot Management from Enterprise Bot Management customers, and describes sampled data. The announcement does not provide a complete plan-by-plan entitlement list for the new views. Check what your zone actually exposes.
A small policy test before a broad change
The following is SysSeder's suggested review process, not a report of hands-on testing.
- Record the starting point. Note the zone, current overall policy, individual agent choices and any relevant application rules. Keep a rollback record. Check actual settings rather than assuming every zone has the same defaults.
- Choose a harmless journey. For example, ask an agent to find one public support article in a test environment. Write down the expected URL and response before the trial.
- Compare access with the intended task. Inspect successful requests, errors and unexpected paths. Where analytics are sampled, avoid treating an absent event as proof that no request occurred. Use appropriate application logs to investigate gaps.
- Keep account permissions explicit. A recognized agent still needs the application's normal authorization checks. Reading a public article, viewing a private account and completing an order should have separately defined permissions and approval requirements.
- Retest after changing the shared policy. Recheck individual choices and repeat the same journey. Confirm the intended result before extending the policy to more agents or more sensitive workflows.
Our reading: document access decisions and verify the outcome. Recognizing an agent does not establish whether its user authorized a purchase, export or account change.