Automated traffic grew 23.5% year over year in 2025 — about eight times faster than human traffic, which rose 3.1%, according to HUMAN Security’s State of AI Traffic report.
AI-driven traffic appears to be a major contributor to that growth, with average monthly volume increasing 187% year over year, while traffic from AI agents and agentic browsers (e.g., OpenAI’s Atlas, Perplexity’s Comet) grew nearly 8,000% year over year.
Automated traffic is defined in the report as: “All internet traffic generated by software systems rather than human users, including traditional automation such as search engine crawlers, monitoring bots, and conventional scraping tools, as well as AI-driven traffic.”
This report follows Cloudflare CEO Matthew Prince’s prediction that bots could overtake human web usage by 2027.
Why we care. Search is increasingly shaped by more than human queries, crawling, and indexing. AI agents now participate in discovery, comparison, and transactions — within Google’s evolving results and across AI-driven interfaces.
The details. HUMAN groups AI-driven traffic into three broad categories:
Training crawlers collecting data for models. They still dominate at 67.5% of AI traffic, but their share is declining as scrapers and agents scale.
Real-time scrapers that feed AI search and answers. Scraper traffic grew nearly 600% in 2025, driven by AI-powered search and real-time answer engines.
Agentic AI systems that execute tasks autonomously. Smaller in share, but growing fastest and most disruptive.
AI agents behave more like users. These systems aren’t limited to reading content. They increasingly navigate funnels, log in, and transact. In 2025:
77% of observed agent activity (requests) occurred on product and search pages.
Nearly 9% touched account-level interactions.
More than 2% reached checkout flows.
About the data. HUMAN analyzed more than one quadrillion interactions (requests/events) across its customer base in 2025, with aggregated, anonymized data from 2022 to 2025. It classified AI-driven traffic into training crawlers, AI scrapers, and agentic AI using user-agent strings, infrastructure signals, and observed behavior, noting limits in self-declared bot identity, which may undercount or misclassify some AI-driven activity.
Bottom line. Traffic is becoming less purely human, and discovery is no longer confined to search engines. Optimization now means deciding which machines can access, interpret, and act on your content.
The report. The 2026 State of AI Traffic & Cyberthreat Benchmark Report