2026-10-06T04:56:04Z https://api.figshare.com/v2/oai

oai:figshare.com:article/34028775 2026-09-30T03:21:32Z category_28864 category_29086 category_28906 item_type_3 month_year_09_2026

AI Agent Page-Type Statistics: Login, Pricing, Documentation and 17 Other Page Types across 40.8 Million Domains, with URL-Rule Results Dr. Samo Plibersek (25092436) Artificial intelligence not elsewhere classified Information security management Data and information privacy AI agents egress filtering Zero trust security Web measurement Web classification Autonomous Agent Cybersecuritization Framework <p dir="ltr">AI Agent Page-Type Statistics is derived from a page-type database that covers 40,786,727 web domains. For each domain the database records which kinds of page it has, such as login, pricing, documentation, careers, legal, help-centre and status pages, together with their verified URLs. These are the pages AI agents act on when they browse the web: they read documentation and pricing pages but can also reach login, signup, upload and checkout pages, so page types are the basis for page-level egress control that allows reads and denies writes for autonomous agents. 16,448,576 domains (40.3%) have at least one verified page type, with 3.61 page types per domain on average; the most common type is contact (70.4% of domains with page types) and 3.26 million domains have a verified login page. The release gives the prevalence of 20 page types, the distribution of page types per domain, page-type prevalence for all 40 site languages and the 60 largest top-level domains, co-occurrence of every pair of page types with conditional shares and lift, and the URL structure of each page type (HTTPS, subdomains, query strings, path depth). It also reports a 40-rule URL layer for agent egress control run on all 58.8 million verified URLs: rules match 0.24% of read-page URLs (false positives) and 45.1% of login URLs, with the first matching rule per page type. A sample of 100 domains with their verified URLs, IAB content categories, web-filtering category, country and popularity group shows the record format; otherwise the release contains counts and shares only. The dataset supports research on AI agent security, zero-trust egress policy, web measurement and multilingual web structure.</p> 2026-09-30T03:21:32Z Dataset Dataset 10.6084/m9.figshare.34028775.v1 https://figshare.com/articles/dataset/AI_Agent_Page-Type_Statistics_Login_Pricing_Documentation_and_17_Other_Page_Types_across_40_8_Million_Domains_with_URL-Rule_Results/34028775 CC BY 4.0