Once ChatGPT enters someone's job, what actually happens to their pay and hours?

This explores what happens to people's pay and jobs once ChatGPT arrives in their work, measured in real-world records rather than predicted. The short answer is that the results depend on where you look. In ordinary salaried employment the effect so far is close to zero. In open freelance markets it is clearly negative.

The clearest evidence of no effect comes from Denmark. National administrative records show that employers adopted chatbots widely and workers took on new AI-related tasks within two years of ChatGPT's launch. Even so, earnings and hours stayed flat within a 2% margin Does AI chatbot adoption change worker pay and hours?. The useful takeaway is the order of events: the content of jobs changes first, and pay and headcount haven't moved yet. A finding of 'no effect yet' doesn't mean 'no effect.' It may mean the change is happening inside jobs, where payroll data can't see it.

Freelance platforms tell a different story. On Upwork, occupations most exposed to generative AI saw employment fall about 2% and earnings fall 5.2% after November 2022, and writing work was hit first and hardest Did ChatGPT's release reduce freelance writing work and pay?. The surprise is who lost most. Lab experiments often find that AI helps weaker workers catch up. In the Upwork data, a strong track record gave no protection, and top freelancers may have been hit disproportionately hard Does a strong track record protect freelancers from AI?. One reason freelancers feel the effect first is that they are the easiest workers for a firm to replace. Firms with more AI exposure replace marketplace contractors with AI tools faster and more cheaply than other firms. That suggests firms that build AI capability in-house gain an advantage that grows with scale, rather than the technology spreading evenly across firms Do firms substitute labor for AI at different rates?.

The pattern also shows up outside paid work. Posting on Stack Overflow fell 16% compared with similar Q&A sites, and the gap grew to about 25% within six months. The drop was largest for programming languages that ChatGPT knows best Did ChatGPT cause Stack Overflow posting to decline?. This isn't an earnings figure, but it measures the same thing: people stop paying for, or asking for, certain kinds of human help when a chatbot can provide it. Usage data also adds context. Most ChatGPT use is now non-work use, and work use centres on writing How is ChatGPT actually being used by real people?, which matches the occupation where freelance losses were largest. Inside companies, adoption is concentrated in larger, R&D-heavy firms and among marketing staff and early-career workers rather than senior staff Who adopts enterprise AI first and how do they use it?.

A note on what the corpus doesn't cover: it has no long-run wage studies of salaried workers and nothing beyond about two years after launch. Read together, these studies suggest a split. Where workers are protected by employment contracts, AI first changes what they do. Where work is bought task by task, it changes how much of that work gets bought. And the losses don't necessarily fall on the least skilled.