Benchmarking Bot Detection Systems Against Modern AI Agents in 2026 (research.poh.org)

A recent benchmarking study revisited and evaluated five leading bot detection systems against modern AI agents and stealth browsers, revealing detection rates ranging from 9% to 99%. Previously, these systems, including Proof of Human and Google reCAPTCHA v3, demonstrated varying effectiveness against automated browsers, but advancements in browser automation and AI have significantly altered the landscape. The current study included a broader array of adversaries, including conventional browser automation and sophisticated AI-powered browsing agents, to assess how well these systems could discern human activity from increasingly capable AI. The findings highlighted a distinct performance gap: while Proof of Human achieved an impressive 99% detection rate overall, systems like Cloudflare Bot Management and hCaptcha struggled, particularly with AI agents that could navigate user interfaces more intelligently. These results underscore the difficulties of distinguishing human users from AI as traditional automation cues become less reliable. As detection technologies evolve, it becomes crucial for future benchmarks to incorporate larger traffic volumes and real human users to refine accuracy and minimize false positives, ultimately advancing security measures in the AI-driven landscape.