Web Hosting

AI Personal Agent Access Policy: Cloudflare’s New Site Owner Controls

A small robot figure, a stand-in for an AI personal agent access policy decision

What happened

As of 2026-11-17 UTC: on November 4, 2026, Cloudflare published a blog post by Marina Elmore and Nicole Justus introducing new analytics and more granular access policies for “personal agents,” the AI assistants that act on a website visitor’s behalf, for example comparing products, finding a support article, or completing a purchase. This article calls the new per-agent options an AI personal agent access policy. That label is ours, not Cloudflare’s.

Why this matters for a site owner

Cloudflare’s post opens with a description of what a personal agent actually is: “the intermediary between humans and the Internet, turning a person’s instructions into requests and actions on websites.” A customer might ask a personal agent to compare products across a site, find support articles, or complete a purchase. Every one of those interactions shows up to the site as automated traffic, and without a clear view of which agents are visiting and what they are doing, Cloudflare says, “it can be difficult to know whether that traffic is helping your business or creating risk.” The post adds that blocking agents altogether can feel like the safest choice, but “it can mean losing customers.”

Cloudflare treats these personal agents separately from the AI bots used for Search and Training, which have their own policies. For background on that other traffic, see our look at the AI bot traffic mistakes site owners keep making. Cloudflare’s post pairs visibility with more granular policies instead of treating blocking as the only option.

What Cloudflare actually launched

Two things shipped together: new analytics to show how personal agents access a site on behalf of humans, and more granular access policies, which this article calls the AI personal agent access policy, to act on that information. Cloudflare frames the goal as letting a site “welcome agents on your own terms, and in ways that support your business,” rather than a blanket allow or block decision.

Seeing which agents are visiting your site

In Security Analytics, the Bot Analysis chart now has a new category called Personal Assistants, shown alongside the existing Verified Bot, Likely Human, Likely Automated, and Automated categories. A new Personal Assistants widget in Top statistics lists which agents are visiting and how many requests each one sends. Filtering on a single agent updates every other widget to show that agent’s paths, status codes, and source networks, so a site owner can see whether an agent is logging in, reading product pages, reaching checkout, or hitting errors on pages it cannot use.

That view is the evidence to check before you set an AI personal agent access policy. Cloudflare says it maintains “the largest verified bot directory in the world.”

Agents and your logged-in users

For sites using Account Abuse Protection, which Cloudflare’s post describes as currently in early access, site owners and fraud teams can see which user each agent represents. A population overview shows which personal agents are active across all users and how many accounts use each one, and every login or sign-up event now records whether a personal agent made the request and which one.

On an individual account’s overview, a fraud analyst investigating an unusual login can check whether it came from an agent the customer has used for weeks or one that appeared for the first time that day. Cloudflare notes that if an attacker appears to be using an agent on a specific account, fraud analysts can mitigate that one account with a WAF rule on its Hashed User ID, without blocking the agent for every other customer who relies on it.

Setting an AI personal agent access policy

Cloudflare says it introduced AI bot policies “this summer” so site owners could make separate access control choices for Search, Training, and Agent behaviors. It is now expanding the existing Agents policy to include granular control options for supported personal agents, and it is renaming the setting “AI Crawler and Agent access policies” to reflect the broader scope.

Inside Security Settings, a site owner can set the overall Agents policy, then choose a different policy for each supported personal agent: Allow (do not block), Block on pages with ads, or Block. In other words, an AI personal agent access policy can differ from one agent to the next.

In Cloudflare’s example, a site owner sets the overall Agents policy to Block on pages with ads, then chooses Allow (do not block) for one supported agent it wants to welcome. Separately, the post notes that one publisher may allow an agent throughout its site while another blocks the same agent on pages where ads support the content. Cloudflare says the existing Search and Training policies stay as they are. For a related Cloudflare feature, see our article on Cloudflare’s bot preference sync and robots.txt.

What happens for new zones

When adding a new zone to Cloudflare, the onboarding flow now asks whether the site monetizes pages with ads and tailors the recommended policies to that answer. Cloudflare says you can always revisit your access controls later in Security Settings, so the onboarding answer is a starting point rather than a lock-in.

What this does not do yet

Cloudflare is explicit that agent identity is only part of the picture. A site may recognize which agent is making a request but lack information it would otherwise have about the person behind it: whether they are an existing customer, what they authorized the agent to do, whether they have a subscription, or which protections should apply.

Cloudflare frames its next goal as helping site owners “preserve the context of their relationships with users when personal agents act as intermediaries.” It says reaching that goal will require privacy-aware ways to carry user context through an agent, help sites assess reputation, and establish clear accountability when something goes wrong. The post says it will share more as it works with agent operators and others across the ecosystem. These are goals, not features available today.

For a site owner, the practical reading is that an AI personal agent access policy tells you which agent is making a request, not what that agent is authorized to do for a specific customer. Only the first of those has a shipped answer in this announcement.

A short checklist for this week

  • If your site is on Cloudflare, open Security Analytics and check the new Personal Assistants category and widget before changing any setting, to see which agents are already visiting and what they request.
  • Choose the AI personal agent access policy per agent, using what the analytics show each one requesting (paths and status codes), not just one overall setting.
  • If ad revenue funds parts of your site, consider Block on pages with ads rather than a full Block, as in Cloudflare’s example of a publisher that blocks an agent on pages where ads support the content.
  • If you use Account Abuse Protection, treat the new per-agent login and sign-up visibility as a fraud-investigation signal, for example an agent appearing on an account for the first time, rather than as a reason to block an agent outright.
  • Do not assume the platform already carries authorization or subscription context through an agent. By Cloudflare’s own account, that is still a goal rather than a shipped protection.

What is still unconfirmed

Cloudflare’s post names Account Abuse Protection as “currently in early access” but does not state which Cloudflare plans include it, a general availability date, or a specific list of which personal agents are “supported” under the new per-agent AI personal agent access policy options. The post says Cloudflare is launching the new analytics “today,” but it does not say how the rollout is staged or whether availability differs by region or plan.

The page’s metadata shows it was modified on November 9, 2026, after its November 4 publication, and this article does not track what changed. Site owners should check their own Cloudflare dashboard to confirm which of these features are visible on their account, and on which plan, before planning around them.

Featured image: “Broken robot toy” by Plutor, licensed CC BY 2.0, via www.flickr.com.