When Patrick Ciriello lost his job and couldn’t find work for nearly a year, his family’s foundation crumbled.“You hear about people who hit rock bottom,” Ciriello told the Guardian.“Well, I was there.”For most of his career, the 60-year-old with a master’s degree in information management designed software systems for banks, universities and pharmaceutical companies.But a series of economic shocks – the dot.

com crash, the 2008 financial crisis and the Covid pandemic – cost him jobs, sometimes forcing him to dip into his savings and retirement funds.Each time, he eventually found another role.That changed at the start of 2023.After losing a job building industrial printer heads, Ciriello sent out hundreds of applications for IT support roles, customer service positions and even a deli counter job at a local supermarket.He didn’t get a single offer.

Even before losing his job, Ciriello, his wife, their 20-year-old son and their cat had been living in motels in northern Vermont for a year, an expense that the state covered because they were experiencing financial hardship.When Vermont ended this funding at the close of 2023, Ciriello’s family had to move out.For about four months, they slept in a Toyota Highlander.They often parked overnight in Walmart lots and spent their days in libraries and McDonald’s with free wifi so he could submit job applications.“I was [practically] getting 1,000 job alerts a day on my email,” he said.

Then, in March 2024, Ciriello got what he calls a “cryptic” message on LinkedIn, advertising a job as a “content writer”.He initially assumed it was a scam.He replied anyway.Only after he started the job did he realize what it actually involved: training artificial intelligence models.Ciriello is one of five skilled workers aged 50 and older who spoke to the Guardian about how, after struggling to find work in their fields, they’ve turned to an emerging and growing category of work: using their expertise to train artificial intelligence models.

Known as data annotation, the work involves labeling and evaluating the information used to train AI models like Open AI’s ChatGPT or Google’s Gemini.A doctor, for example, might review how an AI model answers medical questions to flag incorrect or unsafe responses and suggest better ones, helping the system learn how to generate more accurate and reliable responses.The ultimate goal of training is to level up AI models until they’re capable of doing a job as well as a human could – meaning they could someday replace some of these human workers.The companies behind AI training, such as Mercor, GlobalLogic, TEKsystems, micro1 and Alignerr, operate large contractor networks staffed by people like Ciriello.Their clients include tech giants like OpenAI, Google and Meta, academic researchers, and industries including healthcare and finance.

For experienced professionals, AI training contracts can be a side hustle – or a temporary fallback following a layoff – where top experts can, in some cases, earn between over $180 an hour,But that’s on the high end,For some older workers like Ciriello, it represents another thing entirely: a last refuge in a brutal job market that is harder to stay in, or re-enter, the older they get,For many of them, whether or not they’re training their AI replacements in their professions is besides the point,They need the work now.

US workers over age 60 take about 50% longer to find new jobs than people in their 20s and 30s – and only a fraction regain their previous earning levels, says Richard Johnson, vice-president of the Aarp Public Policy Institute, who researches how age bias impacts job outcomes.He told the Guardian that employers may erroneously view older workers as more expensive, lacking current skills and harder to train than younger people.About half of workers in the US aged 50 to 54 are involuntarily pushed out of long-term jobs before they expect to retire, according to the Urban Institute, a policy research nonprofit.The pandemic only intensified those pressures.Roughly 5.

7 million workers over 55 lost their jobs in early 2020, and many have yet to return to stable work, according to the Economic Policy Institute, also a research organization.“There’s just a lot of desperation out there,” Johnson said.As opportunities narrow, many turn to what Joanna Lahey, a professor at Texas A&M University who studies age discrimination and labor outcomes, calls “bridge jobs” – lower-paying, less demanding roles that help workers stay financially afloat as they approach retirement.Historically, that meant taking temp assignments, retail and fast-food work and gig roles like Uber and food delivery.Now, for skilled workers – engineers, lawyers, nurses or designers, for example – using their expertise for AI data training is becoming the new bridge job.

“[AI] training work may be better in some ways than those earlier alternatives,” Lahey told the Guardian.AI training can offer flexibility, quick income, and intellectual engagement.But it’s often a clear step down.Professionals in fields such as software development, medicine or finance typically earn six-figure salaries that come with benefits and paid leave, according to the US Bureau of Labor Statistics.According to online job postings, AI training gigs start at $20 an hour, with pay increasing to between $30 and $40 an hour.

In some cases, AI trainers with coveted subject matter expertise can earn over $100 an hour.AI training is contract-based, though, meaning the pay and hours are unstable, and it often doesn’t come with benefits.Ciriello’s first AI training job was with a contracting firm that hires workers to train Google’s AI products, including its Gemini model.He made $21 an hour, working 40-hour weeks, to review AI responses and identify errors so engineers could refine how the systems behaved.After about a year, he and his colleagues were let go during a mass layoff in January 2025.

Now he has another AI training role through a staffing firm, where he has evaluated AI model responses for Meta since August 2025.Ciriello works about 40 hours a week earning $20 an hour – far less than he made earlier in his IT career.The income covers rent, car payments, insurance, utilities and food for him and his family, but little else.His earnings are low enough that he qualifies for Medicaid and Snap benefits.“I don’t think I’ll ever be retiring,” he told the Guardian.

Multiple job losses and medical crises over the years have left Ciriello and his family with no savings.What little remains from his paychecks goes toward creating a financial safety net for his son, who is disabled.His wife provides full-time care for him, making Ciriello the household’s sole breadwinner.“I need to set up a trust fund for my son so that when me and his mother aren’t here, he’s taken care of,” he said.More often than not, professionals turn to AI training when they run out of other options.

For more than a decade, Rebecca Kimble, 52, worked as an emergency medicine physician after graduating from the Albert Einstein College of Medicine in 2009.She treated patients and saved lives in emergency rooms across the United States, from Indigenous reservations in South Dakota to a small town in rural Maine, earning between $300,000 and $500,000 a year.“I enjoyed working where other people didn’t like to work,” Kimble told the Guardian.“I wasn’t afraid to go to the middle of nowhere and work with people who didn’t have much.”In February 2022, her career in emergency medicine began unraveling.

First, Kimble stepped away from clinical work for a few months after a DUI placed her on administrative leave.When she was legally cleared to work again later that year, she was diagnosed with breast cancer.The surgery and radiation treatment she underwent kept her away from clinical work for more than two years.When she was ready to return, Kimble expected her experience to carry weight.Instead, she kept getting rejected when she applied to roles.

Extended gaps in practice can make it difficult for physicians to re-enter the workforce, according to the American Medical Association, as hospitals often prioritize candidates who are actively practicing.The realization that she might not return to the emergency room was devastating.“That was absolutely horrid,” Kimble said.“Each time you try and it doesn’t work, you go through that process of fighting depression and anxiety.I thought: ‘Oh my God, this is it.

This part of me died, and I wasn’t ready for it to die.’”Pursuing a master’s degree in public health at Brown University and establishing her own consultancy did little to improve her job prospects.Then her mentor suggested AI data training.By January, Kimble was juggling AI training assignments on several platforms, evaluating how models responded to medical questions.She described the shift as a “phenomenal transition” that allowed her to combine medical expertise with analytical work.

Even so, she would return to emergency medicine if given the chance,“If you give me a job tomorrow that I think I can do, I’ll go do it,” she said,“I miss the ER,”Like Kimble, Anne turned to AI training after her academic career stalled,She asked for anonymity because she’s embarrassed about her financial situation and doesn’t want her kids to know the details.

The 60-year-old, who holds a master’s degree in health sciences and a PhD in public policy, spent 18 years in higher education administration overseeing and teaching in PhD programs.She most recently spent close to two years as an assistant professor, teaching occupational therapy PhD students – a job she found deeply fulfilling.Then she developed long Covid.Brain fog and fatigue made it difficult to teach and keep track of meetings.“I wasn’t functioning at the level that I needed to be,” she said.

Worried that in-person exposure in the classroom was making her sicker, she quit in April of 2023.For eight months, Anne applied to roughly 100 remote roles, searching for positions like “curriculum designer” and “assistant professor”.She received no offers.When she stumbled upon an online job listing for an AI trainer, she was drawn to the promise of learning new skills in a growing industry and applied.Anne went from earning a six-figure salary in academia to getting $26 an hour training Google’s AI models.

She worked 40 hour weeks for nearly two years – with no benefits – before being laid off last September.She now trains Meta’s models through a staffing firm.There, she makes a dollar an hour less than she previously did, just enough to cover her mortgage, student loans and utilities on her own.Anne finds it “personally rewarding” to shape AI to produce better outputs.“I can write a prompt like nobody’s business,” she said.

Still, it is far from what she expected at this stage of her career.“It’s just devastating and demoralizing to think of all the time I spent on my career and the sacrifices I made to earn my graduate degrees,” Anne said.“Look where I’m at now.”The data trainers the Guardian interviewed say the work can be intellectually engaging, but it exposes them to the instability of the gig economy.At his first AI training firm, Ciriello says he and the other workers were required to complete 12 tasks an hour, which included analyzing a prompt, evaluating the response, then rating the model’s accuracy.

He described the environment as a “tech sweatshop”.His current role is more flexible.Anne’s job is also flexible.She doesn’t need to clock in at a specific time and her performance isn’t tied to how fast she can complete a task.The work is consistent.

Despite taking a slight pay cut, she now gets paid for holidays and time off.Kimble’s experience is far less predictable.Rather than fixed shifts, she operates within a gig-style marketplace where assignments appear without warning and are claimed by whichever workers log in first.When she first started, she sometimes woke up at 4am to check for new work.“This is not a job,” she said.

“This is a gig.”Typically, Kimble logs six to nine hours a week doing AI data training.She earns between $500 and $1,000 a month on assignments, with hourly pay ranging from $30 to $140.Sometimes her AI workload can jump to 25 to 30 hours.Other weeks, the work disappears entirely