When I browse internet in 2025, I rarely encounter captcha anymore. There is no italicized text to recognize. There is no image grid of brake lights to identify.
And on the scarce occasion that I’m asked to perform some bot-deterrent task, the experience almost always feels surreal. A friend shared a recent test where they were shown photos of dogs and ducks wearing hats ranging from bowler hats to French berets. The security questions rudely ignored the animal hats and asked us to select photos showing animals with four legs.
Other puzzles are very audience specific. For example, the captcha for Sniffies, a gay sex site, involves users swiping a jockstrap across their smartphone screen to find a matching pair of underwear.
So where have all the captchas gone? And why are the few existing challenges so damn weird? I spoke with cybersecurity experts to better understand the current state of these fading challenges and why the future will likely look even more peculiar.
Bot friction, human frustration
“When captcha was invented, the idea was that it was literally a task that a computer couldn’t do,” says Reid Tatoris, who leads the application security detection team at Cloudflare. The term captcha – a completely automatic public Turing test designed to tell computers from humans – was coined by researchers in 2000 and presented as a way to protect websites from malicious non-human users.
The initial test that most users saw online contained strange characters, usually a combination of garbled letters and numbers, that had to be reproduced by typing them into a text field. The computers couldn’t see what characters they were; people could, even though most of us would have to squint to get it right.
Financial companies like PayPal and email providers like Yahoo have used this iteration to protect against automated bots. More websites eventually added audio readings of the correct answer after pressure from blind and low-vision support groups, whose members were actually people who browsed the Internet but were unable to complete the vision-based challenge.
What if, instead of just a test to stop bots, the challenge could generate actionable data? This was the main idea behind the release of reCaptcha in 2007. Thanks to reCaptcha, users identified words that machine learning algorithms could not read at the time. This accelerated the process of transferring print media to the Internet. The technology was quickly adopted by Google, and reCaptcha played a key role in the company’s efforts to digitize books.
As machine learning capabilities improve – and as people learn to read strange texts – online security checkpoints have become more complex for malicious bots to bypass. The next iteration of the reCaptcha challenges featured image grids in which users were asked to select specific options, such as photos of a motorcyclist. Google used the data collected here to improve your online maps.