OpenAI and Upwork Study: AI’s Effect on Freelance Work
TL;DR: OpenAI’s research with Upwork finds that AI is shifting both what clients hire freelancers to do and how they assemble projects. AI-related freelance postings became more diverse, combined tasks in less familiar ways, and carried larger fixed-price budgets and more hours, while some activities common in ChatGPT use lost share on Upwork—a pattern that does not support a simple substitution story.
What OpenAI and Upwork studied
OpenAI partnered with Upwork to compare client requests on the Upwork marketplace with the activities people undertake directly in ChatGPT. The goal was to detect labor-market changes early: freelance contracts are often short, and clients can adjust requirements with each posting, so freelance demand may move sooner than economy-wide employment statistics.
The stakes are substantial. Upwork’s 2026 Future Workforce Index found that 38% of skilled U.S. knowledge workers freelance. Upwork’s previous annual study estimated that freelance knowledge work generated $1.5 trillion in U.S. earnings in 2024. Those estimates extend well beyond Upwork’s own marketplace.
The research is part of OpenAI’s Work at the Frontier series, in which the OpenAI Economic Research Team explores how AI is changing work in real time.
The headline finding: no simple substitution story
OpenAI’s findings do not point to a simple story in which ChatGPT replaces freelance work task for task. Some activities that were relatively common in ChatGPT use lost share on Upwork, but the pattern was not uniform. AI-related work also spanned a wider mix of activities, combined tasks in less familiar ways, and carried larger fixed-price budgets and more hours of work.
Of the 63 work categories analyzed, 43 followed the pattern in which activities relatively common in ChatGPT lost share on Upwork while less common activities gained share. The remaining 20 work categories showed the reverse. OpenAI states that ChatGPT use alone does not explain changes in freelance demand.
ChatGPT use and freelance demand have different task mixes
The mix of work differed across the two platforms. Writing accounted for a much larger share of ChatGPT use, while video and animation accounted for larger shares of requests for freelancers in Upwork postings.
The post reports several time windows for its comparisons. ChatGPT shares cover January 2025 through June 2026; Upwork shares are for H1 2026. Changes in Upwork share are measured from H1 2025 to H1 2026, in percentage points. A separate scatter comparison uses the natural log of a category’s share of ChatGPT use in H2 2025 divided by its share of Upwork job posts in H1 2023, alongside the percentage change in that category’s share of Upwork postings from H1 2023 to H2 2025.
The comparison captures how people currently use ChatGPT, not the full range of tasks AI can support. OpenAI calls this gap the capability overhang.
Clients are requesting AI in more kinds of work
By late 2025, clients were requesting the use of AI in many more kinds of work. Clients sought freelancers to apply AI in tasks including chatbot development, video production, marketing automation, and application integration.
Within individual projects, AI capabilities were often requested alongside other work. AI-related postings listed about as many tasks as other postings, but they more often combined tasks that had rarely appeared together in earlier jobs.
One example in the post is a project brief that asked for AI-generated video, editing, and motion graphics in a single posting. The freelancer would bring those skills together to deliver a finished video.
AI-related project scope broadened
Between early 2023 and late 2025, the mix of tasks in AI-related postings became 45% more diverse—the equivalent of an increase from about 15 equally common tasks to about 21. The post reports the effective number of tasks as 14.6 in early 2023 and 21.2 in late 2025.
Although posted hourly rates were similar for AI-related and other work, AI-related projects had larger fixed-price budgets and involved more hours of work. OpenAI says this pattern suggests that higher budgets may partly reflect a broader scope of work.
Why freelance data can reveal AI’s labor effects early
AI’s effects on work can be difficult to detect in employment statistics. OpenAI notes that the effects may appear first in the work people do. Because freelance projects turn over quickly, postings are continually refreshed as clients’ needs change. Work within firms can evolve without appearing in a new job posting.
Combined with ChatGPT data, freelance postings offer an early view of which activities people undertake directly with AI, which they hire freelancers to perform, and how AI-related projects differ from other work.
What the findings suggest
OpenAI concludes that AI may be changing both what clients hire freelancers to do and the scope of the projects they post. The findings do not point to simple substitution. Some activities common in ChatGPT use lost share on Upwork, but that pattern was not uniform. AI-related work also spanned a wider mix of activities, combined tasks in less familiar ways, and carried larger fixed-price budgets and more hours.
Together, these patterns suggest that AI may be shifting both what clients seek from freelancers and how they assemble projects.