Cloudflare reports that automated systems now initiate a majority of certain web requests, with 57.4% of requests to a selection of sites classified as bot-originated versus 42.6% human, according to reporting by NBC News and CNET that cite the company's post and CEO Matthew Prince. Prince is quoted as saying, "Welp, that happened faster than I predicted," and noted the rise is driven in part by autonomous AI "agent" systems that crawl many more pages than a human would, per NBC. Separately, a Human Security report cited by CNBC found automated traffic grew roughly eight times faster than human activity in 2025 and highlighted several AI-driven services with rapid growth.
What happened
Cloudflare reports that 57.4% of HTTP requests to a selection of websites it monitors are now automated, while 42.6% are human-originated, per the company post and coverage by NBC News and CNET. CEO Matthew Prince is quoted saying, "Welp, that happened faster than I predicted," in remarks reproduced by multiple outlets, and he told NBC News he was "stunned by the rate of growth." Newser and other reporting also attribute to Prince the comment "It's going to be pay to crawl," describing a possible commercial response to nonhuman traffic.
Technical details
Reporting frames the growth as driven by autonomous AI "agents," defined as systems that programmatically browse and fetch pages on behalf of higher-level services. NBC and CNET describe how an agent can visit thousands of pages during the time a human visits a handful, making agentic traffic disproportionately large per user request. CNBC cites a Human Security State of AI Traffic report that found automated traffic increased far faster than human traffic in 2025, with one example, the crawler OpenClaw, reportedly growing nearly 8,000% year-over-year, per CNBC's summary of the Human Security data.
Editorial analysis - technical context
Industry-pattern observations show that programmatic agents and automated crawlers have long generated substantial machine-to-machine HTTP requests (search-engine indexes, monitoring tools, CDNs). What differs in the current wave is the emergence of more broadly deployed, purpose-driven AI agents that perform multistep web retrieval and synthesis at scale. Observers in the reporting note that these agents amplify request counts because each agent-driven query can expand into hundreds or thousands of fetches, which inflates raw traffic metrics even if end-user engagement signals (time on page, ad clicks) are unchanged.
Context and significance
Editorial analysis: For web infrastructure and monetization, a sustained majority of automated requests alters capacity planning, cost allocation, and measurement baselines for metrics normally assumed to reflect human attention. Reporting highlights geographic and temporal variance in bot share: CNET reproduces Cloudflare region-level data showing some regions and small territories with very high automated proportions, while some countries still show human-dominated traffic. CNBC and other outlets stress measurement challenges, there is no single global corpus of interactions, and cite Human Security's methodology, which processed large-scale telemetry to estimate trends.
For practitioners
Industry context: Observers developing web services, analytics pipelines, or ad measurement should treat raw request counts as increasingly unreliable proxies for human engagement. Typical mitigations in similar situations include refining bot classification, instrumenting higher-fidelity engagement signals (client-side events, authenticated sessions), and separating billable APIs from open crawl endpoints. The published commentary also highlights potential commercial responses discussed publicly, such as differential access or charging for automated crawling, which would change traffic economics if adopted broadly.
What to watch
For readers tracking this shift, monitor vendor documentation for new bot-classification signals from CDN and WAF providers, updates to server- and client-side metrics that distinguish agentic activity, and follow Human Security, Cloudflare, and similar telemetry providers for methodology disclosures. Also watch for any industry standards or publisher policies that attempt to gate or monetize high-volume automated crawling.
Scoring Rationale #
This is a notable infrastructure story with direct operational impact for web engineers, SREs, analytics teams, and ad/monetization platforms. It changes capacity and measurement assumptions but does not introduce a new modeling breakthrough.
Practice interview problems based on real data
1,500+ SQL & Python problems across 15 industry datasets — the exact type of data you work with.