When AI displaces a knowledge worker, does the pay hit fade with time, or does it stick around?

This explores whether the earnings hit to knowledge workers displaced by AI fades or sticks. The short answer is that the collection doesn't yet measure duration. None of these studies follows displaced workers long enough to draw a recovery curve. What it does offer is a sharp early snapshot, some clues about who gets hurt, and long-range scenarios that point toward a lasting shift rather than a temporary dip.

The clearest evidence comes from Upwork. After ChatGPT's release in November 2022, freelancers in the most exposed occupations saw jobs fall about 2% and earnings about 5.2%, and writing was hit hardest at first Did ChatGPT's release reduce freelance writing work and pay?. The unexpected detail is who took the hit. A strong track record offered no protection, and top freelancers may have been hurt most Does a strong track record protect freelancers from AI?. That reverses the lab finding that AI mostly helps weaker workers. It also matters for persistence: if reputation can't shield people, the usual route back to good pay through track record and client relationships may not work either.

Whether a worker recovers seems to depend less on time than on how the exposure is spread across a job. Across firms from 2010 to 2023, when AI touches only a few of a job's tasks, workers can shift into the tasks that remain, and net employment effects stay modest. When exposure is spread across many tasks, there is less room to move Does concentrated AI exposure enable workers to adapt and reallocate?. Firms also don't substitute at the same pace. Heavily exposed firms replace marketplace workers with AI faster and more cheaply, which suggests the pressure builds where in-house AI capability builds Do firms substitute labor for AI at different rates?. Exposure also falls unevenly. In female-dominated occupations it spreads across all wage levels, so lower-paid workers with fewer resources to adapt are exposed too Does AI exposure hit low-wage workers harder in some fields?.

The forward-looking models suggest the gap may be structural rather than temporary. Anthropic's scenarios have average wages rising while knowledge-worker wages stagnate or fall, as more income flows to capital and occupational friction keeps workers from moving into better-paid roles Does AI growth inevitably shift wealth away from workers?. A more extreme model argues that in an AGI economy, wages drift toward the cost of the computing power needed to replace a person, not toward the value that person creates What happens to human wages in an AGI economy?. On that view, "recovery" may not be a meaningful idea.

One adjacent idea complicates the obvious fix of retraining with AI tools. Abilities that AI enhances behave like an exoskeleton: output looks skilled while the tool is present and falls back to baseline without it Does AI assistance build lasting skills or temporary abilities?. Workers who adapt by leaning on AI may not build skills that protect them later. If you're looking for the missing piece, it's longitudinal tracking of displaced workers over years. The corpus has the shock and the theory but not the recovery data.