On Upwork, after ChatGPT arrived, did freelancers' losses fade or last, and did top performers bounce back faster?
How long do ChatGPT employment effects persist for different freelancer groups?
This explores whether ChatGPT's hit to freelance work faded, held steady, or deepened over time, and whether that differed for writers vs. other occupations or for top performers vs. everyone else.
This explores whether ChatGPT's damage to freelance work was a short shock or a lasting shift, and whether some freelancers recovered faster than others. The short answer is that the collection doesn't track how long the effect lasted for each group. What it does have is a clear picture of who got hit, some evidence on whether these shocks fade or grow, and a few mechanisms that suggest the effect may persist.
Start with who got hit. On Upwork, freelancers in the occupations most exposed to generative AI saw jobs fall about 2% and earnings about 5.2% after November 2022, and writing showed the largest *initial* impact Did ChatGPT's release reduce freelance writing work and pay?. The word "initial" matters: it shows the study saw a time pattern, but this summary doesn't say what happened next. The more surprising result is about quality. A strong track record didn't protect anyone. Top-rated freelancers with long work histories were, if anything, hit harder Does a strong track record protect freelancers from AI?. That cuts against lab experiments where AI mostly helps lower-skilled workers catch up. In a real marketplace, the best people may simply have the most to lose.
For how these shocks behave over time, look next door. When ChatGPT launched, Stack Overflow posting fell 16% compared with similar Q&A sites, and the gap grew to about 25% within six months Did ChatGPT cause Stack Overflow posting to decline?. It didn't bounce back. It got worse, and most of all for programming languages ChatGPT knew best. On the supply side, AI-written articles rose for about two years and then leveled off near 50% of new web content Has AI-generated content stopped growing on the web?. Neither is freelancer data, but both point to the same pattern: early effects grow before they settle, and they settle at a new level rather than returning to the old one.
The mechanism notes suggest why freelance losses might last. Clients who leave don't drift back at random: the firms most exposed to AI replace marketplace workers faster and more cheaply, which points to companies building in-house AI ability they won't give up Do firms substitute labor for AI at different rates?. On the worker side, AI turns freelance work from making things into checking AI output, which removes the paid practice freelancers use to keep their skills sharp Does AI turn freelance work into validation instead of creation?. Related research finds that people do better on tasks while they have AI help but don't get better at doing those tasks on their own afterward Does AI assistance help workers learn lasting skills?. Put together, a freelancer who adapts by leaning on AI may win back work now while slowly losing what made them competitive.
One small signal suggests effort still pays off. On Freelancer.com, workers who spent more time editing AI-drafted cover letters were hired more often, even though most barely edited at all Does editing time on AI drafts predict hiring success?. If any group recovers, it may be the freelancers who use AI as a starting point rather than a finished product. The collection doesn't yet test that over time, though.
Sources 8 notes
A difference-in-differences study of Upwork freelancers found that occupations most exposed to generative AI experienced lower employment and earnings after ChatGPT's November 2022 release, with writing work showing the largest initial impact.
An Upwork study found no evidence that past performance or employment history moderated ChatGPT's negative effects on freelancer employment. The data even suggests top freelancers were hit disproportionately hard, contrary to experimental findings favoring low-ability workers.
A difference-in-differences model comparing Stack Overflow to four Q&A sites with lower ChatGPT exposure found weekly posts dropped 16% post-launch, growing to around 25% within six months. The effect was largest for programming languages with more training data in ChatGPT.
Graphite's analysis of 55.4k web articles found AI-primary content rose to 35.9% within 12 months of ChatGPT, reached 48% by month 24, then stabilized near 50% from Q1 2025 onward. The plateau may reflect poor search performance of AI articles.
Higher AI-exposed firms replace online labor marketplace workers with AI tools faster and at lower cost than less-exposed firms, suggesting returns to scale in internal AI capability rather than uniform technology diffusion.
Show all 8 sources
Research suggests generative AI reorganizes freelance labor away from skill-building task completion toward AI output validation. This shift cuts off the paid practice through which gig workers stay competitive, especially compared to salaried employees who receive mentorship and support.
Wu et al. found that workers using generative AI performed substantially better on content tasks, but when performing similar tasks independently afterward, their performance showed no improvement. The capability did not transfer across contexts.
Within workers on Freelancer.com, time spent editing AI-generated cover letter drafts is associated with higher hiring success, even though most workers submit drafts with minimal revision. The paper measured this through click timestamps and application submissions.
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- The Short-Term Effects of Generative Artificial Intelligence on Employment: Evidence from an Online Labor Market
- Signaling in the Age of AI: Evidence from Cover Letters
- From Producing to Validating: How AI Is Deskilling Freelancers
- Generative AI at Work
- Does generative AI narrow education-based productivity gaps? Evidence from a randomized experiment
- Research: Gen AI Makes People More Productive—and Less Motivated
- Are Large Language Models a Threat to Digital Public Goods? Evidence from Activity on Stack Overflow
- Skill Development, Maintenance, Erosion, and Revaluation: How Knowledge Workers Experience Generative AI