Benchmarking Leading AI Agents Against CAPTCHAs

Mathew Hardy, Mayank Agrawal, and Milena Rmus

We test three leading AI models—Claude Sonnet 4.5, Gemini 2.5 Pro, and GPT-5—on how well they solve Google reCAPTCHA v2 challenges.

Benchmarking Bot Detection Systems Against Modern AI Agents

Mathew Hardy and Mayank Agrawal

We evaluate five leading invisible CAPTCHAs and find significant differences in their ability to detect modern bots and AI agents.

Proof of Human. Creating the Invisible Turing Test for the Internet

Mayank Agrawal and Mathew Hardy

AI systems have detectable behavioral signatures that can be used to improve bot detection. Roundtable's Proof-of-Human API verifies proof-of-human invisibly, continuously, and instantaneously.