3 Surprising Ways You Helped Train AI Without Even Knowing – BGR

AI is the future, and it’s gradually transforming all aspects of society, from daily life and business to creativity and science. Whether that is a good or bad thing is up for debate, but one thing is for sure — AI demand is skyrocketing. The companies behind AI are rapidly expanding their data centers to meet this demand, which is something that has led to a widespread shortage of RAM chips that could make your next smartphone cost more. These companies also need quality data to help train their models, and whether you’re aware of it or not, you’re providing it to them for free. Some of the ways in which you are doing this might even surprise you.

You probably know the obvious ways your device usage and online activities are helping train AI. When you interact with public content on social media, you’re feeding the algorithm that determines what to show you next. When you correct Siri, Alexa, and Google Assistant, you’ve helped their developers improve the AI’s speech recognition patterns. Even when you use autocorrect and predictive text in Google Docs or Microsoft Word, you’re teaching the AI patterns that lead to more accurate grammar and spelling suggestions. But you might not know that solving an online puzzle, playing an augmented reality (AR) game, and even reporting spam could also be used to train AI.

A CAPTCHA (Completely Automated Public Turing test to tell Computers and Humans Apart) ensures that only humans can access certain websites instead of bots. It does this through a test that a human can easily solve but a computer would struggle with. Common CAPTCHAs include ticking an “I’m not a robot” check box, identifying distorted text, and selecting specific objects in images. It’s the last option that has revealed a surprising way in which AI companies are training self-driving cars.

Google has its very own version of a CAPTCHA called reCAPTCHA. You have probably seen reCAPTCHA v2, a more advanced, secure, and annoying version of CAPTCHAs, where you’re presented with a photo grid and asked to pick objects like cars, motorcycles, street lights, sidewalks, and storefronts. But according to an X article by @Sharble, the co-founder of an AI-focused marketing agency, those images come straight from Google Street View, and every time you correctly click on the objects, you’re training Google’s computer vision model to identify them in the real world.

A lot of that data goes to Waymo, a company that develops self-driving technology and is owned by Google’s parent company Alphabet. In a blog post released in early February 2026, Waymo announced its self-driving operations will go fully autonomous with the sixth-generation Waymo Driver technology, which will see it expand to even more cities in the United States. You played a part in making this happen, and Google didn’t even have to tell you or pay you.

In July 2016, fans got a dose of Pokémon like never before when Niantic released “Pokémon Go” for Android and iOS devices. It’s a GPS and augmented reality game that sees people venturing into the real world to catch Pokémon. Within two months, over 500 million people downloaded the game worldwide, according to The Pokémon Company. In May 2025, Niantic was acquired by Scopely, a mobile gaming company, which is continuing its development. However, Niantic is not done with “Pokémon Go” completely, and it’s using the game’s data in its AI spin-off company, Niantic Spatial.

In the decade that millions of players have been trying to catch a Jigglypuff or Charizard, they have been helping Niantic Spatial’s AI build an “inch-perfect view” of the world, according to MIT Technology Review. It has used their images to tag landmarks, allowing its AI to create a super accurate model of the world. The AI can use the model to pinpoint locations with near-perfect accuracy using a few pictures of surrounding landmarks and buildings. 

Now this technology will be used in a collaboration with robotics startup, Coco Robotics, to deploy delivery robots across the U.S. and Europe. It will help the robots navigate places where GPS tends to be unreliable, such as high-rises, underpasses, and freeways. At some point in the not so distant future, someone will get their pizza or groceries delivered by a robot, and they’ll have Pokémon Go players to thank for it.

At best, spam is an everyday annoyance that clutters your inbox, but at worst, it’s dangerous and can make you the victim of a phishing scam. When you get an email that is obviously spam, you probably ignore it, delete it, unsubscribe, or send it to the junk folder. However, the best way to handle spam is to report it — nearly every major email provider has this feature by now. This will train the spam filter’s machine learning algorithms to better recognize and block such messages in the future.

An added advantage of reporting spam is that any future emails from that sender will be automatically filtered out. Furthermore, what the AI learns from them can help protect others from email phishing scams. According to a report by the National Consumers League (NCL), phishing and spoofing scams grew by 85.6% between 2024 and 2025, with losses doubling from $1,000 to $2,060. This massive growth was attributed to scammers using generative AI to craft convincing emails, text messages, and even cloned voices. With these malicious actors using AI to trick people, helping improve that same technology to stop them is doing society a service.