Who is actually requesting the web's pages — people, or machines acting on someone's behalf? The charts below are live from Cloudflare Radar, which observes traffic across Cloudflare's global network. Figures in text are a dated snapshot; the charts themselves update on Radar's own cadence.
Share of HTTP requests to HTML content classified as automated (bot) versus human, worldwide. Radar's classification is score-based: "automated" means likely-automated requests observed on Cloudflare's network, and "human" means likely-human requests — not a census of people.
Text alternative — snapshot from Cloudflare Radar, October 15, 2026: of HTTP requests to HTML content, 64.9% were classified automated and 35.1% human.
Chart: Cloudflare Radar (embed, CC BY 4.0).
Share of bot HTTP traffic by the operator behind each bot — the company or project that runs it — worldwide, as classified by Cloudflare Radar.
Text alternative — snapshot from Cloudflare Radar, November 15, 2026, share of bot HTTP traffic by operator over the previous 7 days: Google 23%, Meta 14%, OpenAI 12%, with 21 more named operators and Other (7.3%) making up the rest.
Chart: Cloudflare Radar — Data Explorer (embed, CC BY 4.0). Operator details: Radar Bots directory.
Share of bot HTTP traffic by bot category, worldwide. (A crawler can serve more than one purpose at once; categories are Radar's labels, not a claim about a crawler's ultimate purpose.)
Text alternative — snapshot from Cloudflare Radar, November 15, 2026, share of bot HTTP traffic by category over the previous 7 days: Search 34%, Training 15%, Agent 12%, SEO 9.5%.
Chart: Cloudflare Radar — Data Explorer (embed, CC BY 4.0). More on bot traffic: Radar Bots.
The mix inside AI-bot traffic itself. Every percentage here is a share of tracked AI-bot requests — never a share of all web traffic; no source we have verified supports that wider claim.
Text alternative — snapshot from Cloudflare Radar, October 15, 2026, share of tracked AI-bot traffic: Googlebot 24.2%, ClaudeBot 13.2%, Meta-ExternalAgent 9.1%, Bingbot 8%, Applebot 7.7%.
Text alternative — snapshot from Cloudflare Radar, October 15, 2026, share of tracked AI-bot traffic by declared crawl purpose: Training 43.2%, Training & Search 39.9%, Search 11.6%, User action 4%, Undeclared 1.4%.
Charts: Cloudflare Radar — AI Insights (embeds, CC BY 4.0).
Radar does not measure malicious intent, so this panel is not a Radar chart. It is the static benchmark from the 2026 Thales Bad Bot Report (Thales/Imperva), reporting full-year 2025 figures for all internet traffic observed across Thales's network. It updates only when a new edition of the report appears.
Of all internet traffic in 2025, 53% was automated; 40 percentage points of it was bad bots — malicious automation — and the remaining roughly 13 points benign automation. Human traffic was 47%.
Source: 2026 Thales Bad Bot Report (full-year 2025). Figures as published by Thales; different method and network from the Radar charts above — do not blend the two.
We build tools that intelligent agents can use and people can trust.
Build a web that agents can use and people can trust. We share the evidence and practical patterns to help make it happen.
— agent1st.dev
Agent1st is a small studio where a human lead and a team of AI agents design, build, and run software together.