A study led by researchers from the University of California, Irvine, suggests that bots are significantly better than humans at solving various forms of CAPTCHAs — a line of defense developed to thwart hackers attempting to steal content, embed malicious posts, make fraudulent transactions or slow website traffic.

Dubbed Completely Automated Public Turing test to tell Computers and Humans Apart (CAPTCHA), this line of defense often involves having human users select from a series of images to identify specific objects in those images.

Solving times for various types of CAPTCHAS. Boxes show the middle 50% of participants, and whiskers show the filtered range. Black vertical lines show the median. Source: arXiv (2023). DOI: 10.48550/arxiv.2307.12108

In a series of experiments, 1,400 individuals were tested on different types of CAPTCHAs alongside bots, with researchers suggesting that the bots were both better and faster at selecting appropriate images.

According to the findings, bots outperformed humans in solving various forms of CAPTCHAs including image recognition, puzzle sliders and distorted text. Specifically, bots reportedly solved distorted-text CAPTCHAs nearly 100% of the time whereas humans solved them between 50% and 84% of the time. Meanwhile, humans reportedly took up to 15 seconds to solve challenges while the bots solved them in less than a second.

The study, An Empirical Study & Evaluation of Modern CAPTCHAs, appears in the journal arXiv.