Cloudflare Now Shows You Which AI Agents Are Reading Your Site
Two new controls that decide whether ChatGPT and friends can read your site, plus the one default that most site owners have never looked at.
Cloudflare added Personal Assistants as a category in Security Analytics on 4 November 2026, alongside a Top statistics widget listing which personal agents visit a site and how many requests each sends. The Agents access policy now supports per-agent rules with three actions: Allow, Block, and Block on pages with ads. Personal agents inherit the overall Agents policy by default, and Cloudflare recommends allowing them.1
Every one of those agent visits used to arrive as unexplained automated traffic. A site owner could see that something was hitting product pages and checkout, and had no way to tell a customer's assistant from a scraper. This closes that gap, and it does so on every plan rather than only at the top.2
What is new, and where it lives
| Capability | Where | What it tells you |
|---|---|---|
| Personal Assistants category | Security Analytics, Bot Analysis chart | Agent traffic as a share of all traffic, over time |
| Personal Assistants widget | Security Analytics, Top statistics | Which agents visit, and request counts per agent |
| Per-agent filtering | Any widget, after filtering one agent | Paths requested, status codes returned, source networks |
| Per-agent policies | Security Settings, AI Crawler and Agent access policies | Allow, block, or block only on ad-supported pages |
| Account-level agent view | Account Abuse Protection (early access) | Which accounts' agents are logging in, and which agent |
Prices measured on 27 October 2026. Vendor plan terms, allowances and prices change without notice, so treat every figure on this page as a dated snapshot rather than a permanent fact. Where a number here differs from the vendor's own pricing page, the vendor's page is correct and the figure above is simply older than it.
The filtering is the part that does the real work. Selecting a single agent narrows every other widget on the page to that agent's traffic, so you can see whether it is reading your documentation, reaching checkout and getting errors on pages it cannot use. An agent that arrives, requests four product pages and receives a 403 on all of them is a customer being lost in a way that never shows up in browser analytics.
The policy table, and the default you did not set
Cloudflare's AI Crawler and Agent access policies cover three behaviours: Search, Training, and Agent. The Agent behaviour means automated activity acting in real time on a person's behalf, which includes chat fetch bots and browser-use agents.
Each behaviour takes one of four actions:
| Action | Effect |
|---|---|
| Allow (do not block) | Do not block matching requests |
| Block | Block matching requests across the zone |
| Block on pages with ads | Block only where Cloudflare detects ads |
| Disallow | Training only: tell crawlers not to use your content via robots.txt |
Two details are easy to miss and both matter.
First, each policy applies to verified bots with that behaviour and to matching unverified bots. So a Bot policy is not limited to the polite members of the directory; it reaches the ones pretending to be them.
Second, and more useful in practice, personal agents inherit the overall Agents policy until you set them individually. That means the default is not neutral. If the Agents policy is set to Allow, every supported personal agent is allowed. If it is set to Block, a customer asking their assistant to buy something from your site is turned away until you intervene.
Cloudflare's own guidance leans one way: it recommends Allow for Personal Agents. That is a commercial judgement, not a technical one, but it is the right default for anyone selling anything.
The Block on pages with ads option is the genuinely new idea here. It lets a publisher allow agents across the site but refuse them on the pages where ads pay for the content, which is a policy that did not previously exist as a single setting.
What Free actually gets
The controls are available broadly; the analytics history is where plans diverge. Cloudflare documents Security Analytics retention as 31 days on Security Events and Security Analytics across Free, Pro and Business, rising to 90 days for Security Analytics on Enterprise. The maximum query window is 30 days on Free, Pro and Business against 31 on Enterprise.
In plain terms, a free site can see agent traffic over the last thirty days and can see the last thirty-one days of history. That is enough to answer the question this feature exists to answer: are agents visiting, which ones, and are they getting what they came for.
Account-level analytics are the exception. Those remain limited to Business and Enterprise domain plans, so the view that shows which of your customers are using agents per account is not a free-tier capability.
The Account Abuse Protection screenshots in Cloudflare's announcement, including the per-user agent view and the WAF rule on a hashed user ID, are from an early-access product. Do not plan around those on any plan yet.
Why this is a free-tier story rather than a bot-management story
Blocking AI is the easy, popular, and usually wrong first instinct. The blunt instrument has been available for a while: a robots.txt directive, a managed rule, or an outright block. What was missing was the information needed to block selectively.
An agent that arrives to compare prices and leaves without buying is worth blocking on an ad-supported page and worth allowing on a pricing page. An agent that logs in as a real customer and reads their order history is a customer, and blocking it is a self-inflicted outage. You cannot tell one from the other without per-agent visibility, and per-agent visibility is what shipped this week.
The wider pattern is familiar from the last fortnight of plan changes. Capabilities get added across all tiers, and the tiering lands on the edges: here it is account-level analytics and retention length. That is closer to how a free tier should work than the alternative. Cloudflare's own logging change, effective 1 December, cuts free accounts off at 0.5 GB a day with no option to buy more. This one gives the free tier a control rather than taking one away.
What to check before you change anything
Look at the Personal Assistants widget before touching a policy. A site with no agent traffic has nothing to decide yet, and defaulting to Block on principle costs nothing today and something later, when a customer's assistant is the only thing trying to reach your checkout.
If agent traffic is present, filter by agent and read the status codes. A well-behaved agent receiving 200s on product pages is doing your marketing. One receiving 403s is broken, and the fix is usually on your side, not theirs.
Then decide per agent rather than globally. The four actions are coarse, but they are the difference between a policy you can defend and a robots.txt directive that crawlers are free to ignore.
Frequently asked questions
What are personal agents in Cloudflare analytics?
AI assistants acting for a person in real time, such as chat fetch bots and browser-use agents. They now appear as a Personal Assistants category in the Bot Analysis chart.
Are agent policies available on the free plan?
Yes. The policy controls and 30-day Security Analytics query window are documented for Free, alongside 31-day retention.
What is the default policy for personal agents?
They inherit the overall Agents policy until set individually. Cloudflare recommends the Allow action for personal agents.
What does Block on pages with ads do?
It blocks the agent only on pages where Cloudflare detects advertising, letting it through everywhere else.
Do these policies apply to unverified bots?
Yes. Each policy covers verified bots with the matching behaviour and matching unverified bots.
Can I see agent activity per account?
Only with Account Abuse Protection, currently in early access, and account-level analytics are otherwise limited to Business and Enterprise domain plans.
Personal Assistants category, Top statistics widget, per-agent filtering, the renamed AI Crawler and Agent access policies, the three actions, inheritance by personal agents and the Allow recommendation verified in Cloudflare's 4 November 2026 announcement and its AI Crawler and Agent access policies documentation. Retention and query-window limits by plan from Security Analytics documentation, all read on 5 November 2026.
Sources
- Cloudflare Blog: personal agent visibility and control, 4 November 2026: blog.cloudflare.com/personal-agents-visibility-control. Used for: Personal Assistants category and widget, per-agent filtering, the renamed policy setting, the action list and personal agent inheritance.
- Cloudflare Docs: AI Crawler and Agent access policies: developers.cloudflare.com/bots/additional-configurations/block-ai-bots. Used for: The three behaviours, the four policy actions, verified and unverified bot coverage, and the Allow recommendation for personal agents.
- Cloudflare Docs: Security Analytics: developers.cloudflare.com/waf/analytics/security-analytics. Used for: Retention of 31 days and a 30-day query window on Free, account-level limits, and Enterprise retention.
All sources are the vendor's own plan or documentation pages, checked on 27 October 2026. Plan terms change without notice, so confirm current limits before you rely on them.
