Fixing AI’s mistakes is becoming a growing side hustle for college grads and freelancers

Since the launch of ChatGPT two years ago, AI has reshaped how we live, learn, and work—touching almost every part of daily life. But while AI appears to have answers to almost every question, it’s far from flawless. From clunky answers to bizarre hallucinations, AI models still rely on human judgment to get things right. That need is turning into a surprising way to make money: fixing AI’s mistakes. Across the U.S., freelancers, especially college grads, are earning by teaching AI to think more like humans.

For years, people have worried about AI replacing jobs. But it turns out that AI needs humans more than we thought. Behind every polished AI response is a person refining its accuracy, fixing hallucinations, and making sure its answers actually make sense.

This isn’t about losing jobs to machines—it’s about jobs changing shape. AI is creating new kinds of work where human skills like judgment, creativity, and expertise matter. The future might not be a battle between humans and machines but a collaboration where each strengthens the other.

Fixing AI’s Wrong Answers Is The Growing Side Hustle For College Grads and Freelancers

That’s where this side hustle comes in. Platforms like Outlier, owned by Scale AI, have opened doors for people to help train AI models. It’s about correcting mistakes and guiding AI to respond in a way that feels natural. What used to be a technical problem is now a new way to earn from home. As Forbes reports, the growing demand for freelancers to fix AI mistakes is quickly becoming one of the hottest gig opportunities.

For example, Scott O’Neal from Covington, Louisiana, juggles two completely different jobs. By day, he handles plumbing sales, which are especially busy when cold snaps freeze pipes. “It’s been super busy—freeze-broken pipes,” he told Forbes. But by night, O’Neal tackles AI, rating and correcting responses for Scale AI, a company valued at $14 billion. His work helps models like ChatGPT produce better answers.

How AI Training Really Works

O’Neal’s tasks change depending on the day. Sometimes, he checks if an AI’s response is accurate and natural-sounding. Other times, he chooses the better of two AI-generated answers or rewrites them entirely if both fall short. With his background in web development, he earns between $300 and $1,000 a week, depending on how many hours he puts in.

He’s one of many freelancers working on Outlier. The platform connects people with paid tasks to help train AI models for companies like Google, Meta, and OpenAI. Outlier launched in 2023, soon after ChatGPT ignited global excitement about AI.

Why AI Still Needs Human Input

AI models like Google’s Gemini or Meta’s Llama can generate everything from email drafts to coding solutions. But their ability to sound smart depends on hours of human fine-tuning. People like O’Neal rate AI responses, filter out inappropriate content, and check for accuracy.

As AI systems become more advanced, they still rely on people with specialized skills. Scale AI says 87% of Outlier’s contractors have a college degree. Of those, 48% hold bachelor’s degrees, 27% have master’s degrees, and 12% hold PhDs. These freelancers bring expertise from creative writing to advanced math and high-level coding.

Why Scale AI is Focused on U.S. Workers

Scale AI is betting big on U.S. talent. Instead of outsourcing, it has built a contributor network across 9,340 towns in the U.S., including major cities like Houston, Chicago, Los Angeles, New York City, and Atlanta. But the reach doesn’t stop there—smaller towns like Rexburg, Idaho, and Lake Mary, Florida, are also part of the mix. About 19% of Outlier’s contractors are based in rural areas.

Scale pitches Outlier as the next step in gig work, offering flexibility like driving for Uber or delivering with Postmates but from the comfort of home. Uber itself even launched its own data labeling platform, Scaled Solutions.

Challenges Behind the Growth

Scale AI isn’t without criticism. Despite reaching $1 billion in annualized revenue as of September, the company has faced backlash. Over the past few months, at least three lawsuits have been filed by Outlier contractors, accusing Scale of poor working conditions, lack of mental health support, and wage theft. These are common concerns raised against tech companies that rely heavily on contract workers.

Still, the demand for human oversight in AI isn’t going away. AI may be smart, but it still needs people to guide it. And for those looking to make some extra cash, fixing AI’s mistakes could be one of the most unexpected side hustles to emerge in recent years.

For an in-depth analysis, check out Richard Nieva’s video below.