← Back to feed

On April 27, 2026, bots crossed 50% of all global HTML requests on Cloudflare’s network — the first time in the internet’s history that automated traffic outnumbered human traffic. As of early June, Cloudflare Radar shows the split at 57.5% bots, 42.5% humans.

In the US, the gap is wider: 71.5% of domestic web requests are bot-generated.

Cloudflare CEO Matthew Prince posted on X on June 3: “Welp, that happened faster than I predicted. Thought it would be end of 2027, then early 2027, but agentic traffic growing so fast that bots have now passed human traffic online for the first time in the Internet’s history.”

Prince had walked through the mechanism at SXSW earlier this year. A human shopping for a product visits five websites. An AI agent performing the same task queries 5,000. Multiply that across hundreds of millions of agentic sessions running round the clock, and the arithmetic works out fast.

The Numbers Behind the Shift

Cloudflare serves approximately one-fifth of all websites worldwide, making its Radar dashboard the largest public window into bot-versus-human traffic composition. The Imperva Bad Bot Report, an independent source, independently confirmed automated traffic crossed 50% for the first time in a decade, reaching 51% of all global web traffic in 2024 — a year ahead of Cloudflare’s internal projections.

AI-driven traffic specifically surged 187% in 2025, growing roughly eight times faster than human web activity during the same period.

Of the 57.5% bot share, not all of it is agentic AI traffic. Cloudflare’s breakdown:

• 37% of automated traffic: malicious “bad bots” (scrapers, credential stuffers, account takeover tools)
• 14%: legitimate crawlers (search indexers, monitoring agents)
• The remainder: AI training crawlers and real-time agentic queries

The top AI crawlers by request share in Q1 2026: Googlebot (31.6% of AI bot traffic), Meta-ExternalAgent (16.7%), GPTBot (12.0%), ClaudeBot (11.7%), Bingbot (8%).

Training crawlers — bots that exist solely to harvest content for model pretraining — now constitute 49.9% of all AI bot traffic, reaching the projected Q2 milestone one quarter early.

What This Breaks

The entire economics of the web were built on the assumption that a human loads a page, sees an ad, clicks a link, fills a cart, and generates signals that publishers, stores, and SaaS companies can monetize.

AI agents break every link in that chain. They read pages on behalf of humans while bypassing ad impressions, session time, scroll depth, click behavior, and conversion events — the four pillars that funded digital advertising for two decades.

CPM and CPC models price against human attention. When bots generate the majority of requests, traffic dashboards reflect machine behavior instead of genuine audience engagement. Conversion rates collapse not because the product deteriorated but because the denominator is no longer a population of buyers.

Analytics built before the agentic wave cannot distinguish a human visitor from a Claude-powered research agent running 3,000 parallel page loads. For publishers, the practical consequence is that reported traffic can rise while monetizable human attention falls.

The Response

Cloudflare has already begun blocking AI crawlers by default unless they compensate content creators — an early implementation of pay-to-crawl protocols that are gaining traction across the CDN industry.

Publishers and media companies are adding robots.txt rules targeting specific AI crawlers, though the Presenc AI analysis of Q1 2026 Cloudflare data notes that if a site only has rules for search crawlers, it is addressing roughly half of the relevant traffic.

The web Prince described at SXSW — “a human typed a query, a search engine found the page, the person clicked through” — has already ended. What comes next is an infrastructure question: whether the sites AI agents depend on can stay funded without the human attention that previously paid for them.