For years, the biggest fear around artificial intelligence was that it would replace workers altogether. However, the reality emerging from a recent data looks different. Instead of triggering mass layoffs, AI appears to be squeezing paychecks.
A new white paper by Apollo Global Management suggests the technology’s biggest impact so far has been on wages rather than employment. Workers in jobs with the highest exposure to AI have seen slower wage growth since generative AI entered the mainstream in 2023, even as overall employment in those occupations has remained largely unchanged.
The raise that never came
The study tracked wage and employment data across 321 occupations in the US and found that jobs with the highest AI exposure recorded an average 6.7% decline in real wage growth after 2023, when ChatGPT became widely adopted. Regardless of fears that AI would replace workers, the researchers found no evidence of widespread job losses.
“Yet, AI’s effects on overall employment were not ‘detectable,'” Sania Edlich, an analyst, and Torsten Sløk, Apollo’s chief economist, wrote in the report. The finding points to a different kind of disruption. Instead of eliminating workers, companies appear to be using AI to improve productivity while keeping headcount largely intact; the gains are shown in slower wage growth rather than layoffs.
The lowest-paid workers are taking the biggest hit
According to the study, service workers have experienced a 24.3% decline in earnings growth since 2023. Workers in the bottom 25% of earners have seen wages fall 10.7% over the same period. By comparison, the highest-paid workers showed no “significant effect” on their wages.
The findings suggest AI is widening existing income gaps, with lower-income occupations bearing the brunt of the adjustment while higher-paid professionals remain relatively insulated.
Some jobs are already feeling the pressure
The impact is already visible in several AI-exposed occupations. Computer programmers saw real wages decline 6.1% between 2022 and 2024, while statistical assistants recorded a 5.4% drop. Software quality assurance analysts and testers, database architects, medical transcriptionists and wholesale sales representatives also reported lower real wages over the period.
Not every decline can be linked to AI alone. Broadcaster announcers and radio DJs, despite relatively low AI exposure, saw wages plunge 52% over two years, reflecting broader industry challenges.
At the same time, AI exposure has not prevented wage gains everywhere. Personal finance advisors posted an 8.4% increase in real wages, while administrative law judges, adjudicators and hearing officers saw wages rise 17.5% despite moderate AI exposure.
Apollo estimates that around 5.8 million US workers currently hold jobs with high AI exposure. The researchers believe that number is only likely to grow as companies embed AI deeper into everyday work.
The study analysed Bureau of Labor Statistics data alongside Anthropic’s Economic Index, which measures how much of a job’s tasks can be performed using AI tools. The debate around AI is no longer just about whether machines will replace people. Increasingly, it is about whether workers can expect the same pay for jobs that AI is making faster, cheaper and more productive.