The strange case of the strange and disappearing captcha
As I browse The Web in 2025 I rarely encounter verification codes. There is no italic text to highlight. There is no photo grid of stop signs to identify.
On the rare occasions that I’m asked to complete some robot-deterring mission, the experience always feels surreal. A colleague shared recent tests in which he was shown pictures of dogs and ducks wearing hats, from bowler hats to French berets. The security questions rudely ignored the animal hats and asked them to choose pictures that showed four-legged animals.
Other mysteries are very specific to their audience. For example, captcha for Sniffies, a gay hookup site, prompts users to swipe a jockstrap across their smartphone screen to find a matching pair of underwear.
So, where did all the verification codes go? And why are the few challenges that exist so strange? I spoke with cybersecurity experts to better understand the current state of these vanishing challenges and why the future may look even stranger.
Robot friction and human frustration
“When CAPTCHA was first invented, the idea was that this was a task that a computer literally couldn’t do,” says Reid Tatoris, who leads the application security detection team at Cloudflare. The term captcha — a fully automatic public Turing test to distinguish between computers and humans — was coined by researchers in 2000 and introduced as a way to protect websites from malicious, non-human users.
The initial test that most users have seen online contains funky characters, usually a set of distorted letters and numbers that you have to repeat by typing them into a text field. The PCs couldn’t see the characters; Humans can do it, even if most of us have to squint to get it right.
Financial companies like PayPal and email providers like Yahoo have used this redundancy to ward off automated bots. More websites eventually added audio readouts of the correct answer after coming under pressure from advocacy groups for the blind and visually impaired, whose members were humans browsing the web but were unable to complete the vision-based challenge.
What if this challenge, instead of just a test to keep bots out, could generate useful data? That was the basic idea behind reCaptcha in 2007. Using reCaptcha, users identified words that machine learning algorithms couldn’t read at the time. This accelerated the process of transitioning print media to an online form. Google quickly acquired this technology, and reCaptcha was instrumental in the company’s efforts to convert books into digital format.
As machine learning capabilities improve – and they learn to read unconventional text – online security checkpoints have adapted to become more difficult for malicious bots to circumvent. The following iterative reCaptcha challenges involved grids of images where users were asked to select specific options, such as images containing a motorcyclist. Google used the data collected here to Improved its online maps.