Why Clicking I'm Not A Robot Isn't Always Safe: CAPTCHA's History, Evolution, And The ClickFix Threat
While CAPTCHA has evolved from simple text and puzzles to behavioural verification, fake CAPTCHA scams now exploit users to deliver malware through ClickFix attacks.
Published : July 7, 2026 at 5:39 PM IST
Hyderabad: Most people would say being a human means having empathy or compassion. But sometimes, it is simply about being able to identify traffic lights, bridges, or vehicles on a computer screen.
In the digital world, being a human is only a matter of not being a bot. The authentication system that proves that is called CAPTCHA. It stands for Completely Automated Public Turing test to tell Computers and Humans Apart and does exactly what its name suggests—validate whether a user is a human or a bot.
CAPTCHA may exist as random strings of text that you need to reproduce or simply click on a checkmark that says, “I’m not a robot”. Apparently, even the most advanced artificial intelligence systems, bots, and agents struggle to do so and have been documented to enlist a human’s help in passing the authentication.
The reason why they exist is to stop cybercriminals from using bots to access a website or online forms for malicious purposes. However, as creative as bad actors tend to be, they have ended up utilising the public familiarity with CAPTCHA as a means to spread malware. Some call them Fake CAPTCHA scams, while the cybersecurity industry calls them ClickFix.
Before heading into the intuitive ways malicious actors have discovered to keep the scam-mill running, let’s first understand the CAPTCHA and its evolution.
The birth of CAPTCHA and the need to verify if you are a human
Several different groups developed the earliest forms of CAPTCHA technology in parallel during the late 1990s and early 2000s to combat hackers that used automated bots for nefarious activities on the internet. For example, computer scientists working for the search engine AltaVista wanted to stop bots from adding malicious web addresses to the company’s link database.
The first CAPTCHA-like system was filed by researchers at the IT company Sanctum in 1997, while the term "CAPTCHA" was coined in 2003 by researchers at Carnegie Mellon University, led by Luis von Ahn and Manuel Blum.
These Classic CAPTCHAs asked users to identify distorted letters or numbers and type them correctly to prove they are human. The distortion makes it difficult for bots using optical character recognition (OCR) to read the text, while humans, who are used to seeing and interpreting letters in different handwriting styles, can usually recognise it. Companies like Yahoo quickly adopted the technology to stop spambots and prevent fake account creations.
CAPTCHAs are being used to date for website gate actions, such as submissions, logins, comments, ticket purchases, and account creation. Koushik Pal, Threat Intelligence Researcher, CloudSEK, explains that it blocks credential stuffing, scraping, spam, ad-fraud clicks, and scalping bots.
Adding further, he says, “It's an access-control or anti-automation layer, not really ‘security’ in the encryption sense. It is more like rate-limiting for behaviour.”
Today, automated bots generate a significant share of internet traffic, with some performing useful tasks like search engine indexing while others carry out harmful activities such as phishing, spam, fake account creation, password attacks, misinformation, and DDoS attacks. As AI systems become capable of solving traditional CAPTCHA challenges, websites are adopting more advanced verification methods to prevent abuse.
The evolution of CAPTCHAs
Back in the 2000s, as CAPTCHAs cemented their role as human-verification systems for website gate actions, the technology continues to evolve in response to bots becoming more sophisticated.
In 2007, Luis von Ahn introduced reCAPTCHA v1, which not only made text challenges harder but also helped digitise scanned books by asking users to identify words that OCR software could not recognise. Google acquired reCAPTCHA in 2009, using it for projects like Google Books while offering it to websites worldwide.
By 2012, Google replaced distorted text with image-based challenges, asking users to identify objects such as traffic lights or taxis. These image CAPTCHAs were more resistant to OCR-based attacks and worked better on mobile devices.
Throughout the years, the CAPTCHA has existed in many forms, across text, image, audio, video, puzzle, and games.
As advances in machine learning gave more power to bots, it prompted the development of more advanced CAPTCHA systems like behavioural verification methods (reCAPTCHA v2). When you click the “I’m not a robot” checkbox, reCAPTCHA does not just look at the click but analyses a wealth of background information to verify the “humanity”. One of these techniques is analysing micro-movements like the mouse cursor or reviewing browser cookies.
Cloud infrastructure platform Cloudflare explains that the test tracks the movement of the mouse as you move it towards the checkbox. While a bot moves the cursor in perfectly straight, precise lines, human movement naturally contains tiny, unconscious microscopic imperfections and randomness. The difference between bot precision and human unpredictability helps reCAPTCHA v2 determine the user.
Another method entails the system reviewing browser history and cookies, analysing the recent digital footprints to distinguish a human from a bot. If the history shows normal human activity, such as watching cat videos, checking your Gmail, or interacting on social media, the system gains confidence that you are a real person.
Talking to ETV Bharat, Koushik Pal explained how the integrity of the wall against bot activity varies depending on the type of CAPTCHA. He said that the “type” mostly reflects which generation of anti-automation signal it's built on, not just visual difficulty.
He said that text-based distorted-character (gimpy-style) CAPTCHA is the weakest today. “Commercial OCR and ML solvers, plus human CAPTCHA-farms, break these cheaply,” he added.
Pal ranked reCAPTCHA v2 (checkbox) better, as it relies on behavioural telemetry, such as mouse movement, cookies, and browsing history, more than the puzzle itself. He ranked reCAPTCHA v3/ Cloudflare Turnstile even higher. “It is invisible, score-based, and analyses background signals continuously. It is harder to automate against,” he said.
ClickFix: Fake CAPTCHA Scams
As websites continue to rely on CAPTCHA systems to protect themselves, users have grown accustomed to them. Cybercriminals now want to exploit this familiarity using fake CAPTCHA, designed to inject malware on a user’s system. We came across several of these, masquerading as Cloudflare security checks and Google's reCAPTCHA, asking us to run a command on the system in the guise of proving we’re human. A dissection of the copied command revealed the hacking attempt.
When asked about these scams, Koushik Pal says, “It’s dangerous for the uninitiated. This is a named technique in threat intel now, called ClickFix.”
Explaining the nature of the scam, Pal says that it skips exploiting software vulnerabilities entirely and instead socially engineers the victim into running the malicious command themselves via the Run dialogue or PowerShell, which relies entirely on manipulating user behaviour rather than exploiting software flaws.
🚨 Came across another malware-injection attempt, this time disguised as #reCAPTCHA.
— Faisal (@itsmeFSL) June 3, 2026
It automatically copied a command, which (long story short) likely connects to a remote server and uses Windows rundll32 Utility to execute malicious code from a DLL (Dynamic Link Library) file. https://t.co/safr01kWqF pic.twitter.com/Xbm1MxLwo0
“This self-execution is what makes it evasive as it grants attackers remote access, enabling credential theft and financial fraud,” he adds.
“By March 2025, ad-based delivery of this technique had more than tripled month over month, with over 600,000 attempted attacks blocked across hundreds of sites,” Pal states. “Typical payloads are information stealer malware and RATs, capable of harvesting saved passwords, banking data, and browser session cookies.”
In some cases, the ClickFix also appeared on some small-time blogs, which ran as proper informative websites on a particular niche for years, prompting doubts about whether the website owners were running the fake CAPTCHA or they were tricked into hosting the ClickFix.
Koushik Pal says that almost always, the site owners do not know they are being used to spread malware. “They are usually a secondary victim, not a co-conspirator,” he adds.
Worth repeating: a legitimate reCAPTCHA should never ask you to open Run (Win+R), Command Prompt, PowerShell, Terminal, or paste commands into your computer.
— Faisal (@itsmeFSL) June 3, 2026
If it does, you're not verifying you're human—you're being tricked into handing control of your system to someone else.
Pal further explains that bad actors target site owners using two common mechanisms. The first one is malvertising, where a malicious ad, served through a legitimate ad exchange, injects the fake CAPTCHA overlay. In this case, the publisher never even touches the code. The second one is CMS/plugin compromise, where attackers exploit a vulnerability to steal admin credentials or API keys and silently inject the JavaScript themselves.
Letting the website run ClickFix could be one of the fastest ways to lose your site reputation, so the website owners need to scrutinise both the front- and back-ends of their site. Users visiting the website need to remember that genuine CAPTCHA never asks you to run a command on your PC. It purely stays within the browser and at most can ask you to “Select all squares with Traffic Lights”.
