Bots are now better and faster than humans at passing Captcha tests, according to a new study by researchers at the University of California, Irvine.

Captcha (which stands for Completely Automated Public Turing test to tell Computers and Humans Apart) tests have become common enough across different websites to determine if users are human or robots – you often see them on sites such as Ticketmaster to prevent users from employing bots to sweep up tickets.

With the advent of artificial intelligence, bots are becoming more and more capable to solve complex problems and interpret data, making it more difficult to Captcha tests to pick them out.

Researchers tested different Captcha systems from across over 100 popular websites, all ranging in complexity and type.

Then, with 1,000 human candidates varying in age, sex, education, and location, each took ten different Captcha tests.

Researchers then compared their performance against that of different bots which had been developed by researchers.

It was the bots that did better, according to the study, which is awaiting its peer review.

Human accuracy ranged from around 50-84%, while the bots accuracy ranges from 85-100% across the Captcha tests.

Bots also beat humans in terms of speed. In a distorted text Captcha test, humans took about nine to 15 seconds to complete the challenge, while bots needed less than a second and passed with 99.8% accuracy.

The only type of captcha test showing similar accuracy levels between bots and humans were the reCAPTCHA-image test – think the ones where you have to select picture squares that include a stoplight, or a zebra-crossing. Humans had an accuracy of 81%, but bots still outperformed them, with an accuracy of 85%.

The bots are clearly beating humans at their own game.

Recommended reading

• Malicious Bots Had a Record Year in 2021
• Twitter Bots Responsible for a Quarter of Climate Crisis Posts
• Two Thirds of Internet Traffic is Now Made Up of Bots

The new research brings into question the effectiveness of Captcha tests, and if it is time to bring in entirely new methods to tell the different between AI and humans.

As AI becomes more mainstream and accessible, the need to police its use also increases as threat actors can use the technology to bypass security systems.

To combat the issue of bots taking over, Andrew Searles, a researcher with the UC Irvine team, suggested the use of background algorithms to detect bots on websites rather than relying on an outward, user-facing test, he told the New Scientist.

Once bots can accomplish a task, they generally become faster at completing it, the research says, meaning developers are likely already behind in preventing bots from bypassing their Turing Tests.