Most Workers Say AI Hasn't Changed Their Job. The Numbers Tell a Different Story.

The headline from a FlexJobs survey released this month is, on the surface, reassuring: 75% of U.S. workers say they have not experienced any job changes due to AI. Nearly half say they don't use AI tools at all. If you're looking for evidence that AI displacement fears are overblown, this survey is a reasonable place to start.

Look a little deeper and the picture gets more complicated.

Workers aged 22 to 25 in occupations with high AI exposure have seen a 13% decline in employment since 2022, according to Dallas Federal Reserve research. Entry-level hiring at the top 15 technology companies fell 25% between 2023 and 2024. The FlexJobs survey also found that 42% of workers are worried about AI-driven displacement, even among the 75% who say they haven't felt it yet. That gap between "hasn't happened to me" and "I'm worried it will" is doing a lot of work.

The pattern that emerges from the data looks something like this: most workers in established roles have not seen their jobs change yet, but the front edge of the disruption is hitting specific categories hard. Administrative roles face the highest exposure, with 26% of positions at direct risk, followed by customer service at 20%. Young people entering the job market in 2024 and 2025 are finding fewer entry-level positions available, not because companies are in financial trouble, but because AI is handling what those positions used to cover.

There is a counterargument worth taking seriously. J.P. Morgan research shows AI job postings in the U.S. running 134% above 2020 levels, with 275,000 positions requiring AI skills advertised in January 2026 alone. The displacement story and the creation story are playing out simultaneously, which makes the net effect hard to assess in real time.

Washington University's labor market research found that whether workers adopt AI depends heavily on whether they see it as a career-relevant learning tool rather than a productivity shortcut. Workers who treat AI as something to master are probably better positioned than the aggregate figures suggest. Anthropic's labor market research, released this year, found meaningful variation by industry and occupation that broad surveys tend to flatten.

What the evidence actually suggests is this: the 75% figure is real, but it reflects a survey population weighted toward workers in roles that AI has not yet targeted most heavily. The people most at risk are younger, in more routinizable positions, and may not show up proportionally in self-selected online surveys. The rising anxiety among workers who say they haven't been affected yet suggests that awareness of second-order effects is spreading faster than the first-order impacts themselves.

Whether this is a slow structural shift or a temporary disruption with a long adjustment period is a question nobody can answer definitively right now. The honest reading: not a crisis for most workers today, but a real and measurable change for a specific group, with pressures that have not fully played out.