What happened
OpenAI’s November 10, 2026 announcement presents research conducted with Upwork, an online marketplace connecting clients and freelancers. The researchers compared requests in Upwork postings with activities people undertake directly with ChatGPT. The company frames freelance work as an early signal because contracts are often short and client requirements can change with each posting. The announcement links a key-takeaways PDF and a full report.
The comparison covered 63 matched work subcategories. Across those categories, 43 showed a pattern in which activities relatively common in ChatGPT use lost share in Upwork postings, while less common activities gained share; the remaining 20 showed the reverse. OpenAI says this does not establish that ChatGPT use caused changes in freelance demand. The comparison uses ChatGPT activity from January 2025 through June 2026 and Upwork posting data over specified periods, including changes from the first half of 2025 to the first half of 2026 in one chart. The report also notes a “capability overhang”: observed ChatGPT use does not capture every task AI could support.
The announcement says clients were requesting AI in a wider range of work by late 2025, including chatbot development, video production, marketing automation, and application integration. AI-related postings contained an average of 3.3 tasks, about 5% more than the 3.2 in other postings. From early 2023 to late 2025, the mix of tasks in AI-related postings became 45% more diverse, equivalent to moving from about 15 to 21 equally common tasks. Among jobs with at least two tasks, an average 8.4% of task pairs in AI-related postings had rarely appeared together historically, compared with 2.4% for other jobs.
For AI-related projects, adjusted fixed-price budgets were 52.4% higher and hours during the first 90 days of hired hourly contracts were 40.6% higher than for comparable non-AI work. Posted hourly rates were 1.5% higher. These comparisons were adjusted within the same type of work and posting month; budget and hours data run through April 2026. OpenAI says larger budgets may partly reflect broader project scope. The announcement does not describe a product, pricing change, or new tool access.
What the archive adds
New Runtime’s earlier coverage of OpenAI’s work data described quarterly figures on ChatGPT use across 165 work activities and nearly 100 countries, covering Q1 2025 through Q3 2026: ChatGPT’s expanding world of work. This is a new fact on a known story: the present research pairs observed ChatGPT activity with marketplace postings and project characteristics.
Trend signal
This supports the broader shift toward organizing work around specifying, routing, verifying, and owning outputs, rather than treating AI as a simple productivity multiplier. The evidence is suggestive rather than conclusive about changes inside companies, because it measures freelance postings and ChatGPT use. See AI-native organizations move toward review and orchestration.
Why now
OpenAI argues that freelance contracts turn over quickly and that postings are continually refreshed, making them a potentially earlier indicator of changing client requirements than economy-wide employment statistics. The study combines this marketplace view with ChatGPT activity to examine both tasks people do directly with AI and tasks they hire freelancers to perform.
Why it matters
For teams commissioning AI work, the findings caution against budgeting around a narrow automation task when the actual deliverable may span generation, editing, integration, and review. The example of a video brief combines AI-generated video, video editing, and motion graphics; the measured rise in uncommon task combinations points in the same direction. Higher fixed budgets and more hours do not by themselves show that freelancers earn more per hour or that AI caused the difference. They may indicate wider scopes. Product and operations teams can use this as a prompt to define the full outcome, dependencies, and acceptance criteria before estimating savings or staffing. It also suggests that evaluating AI adoption only by which isolated tasks disappear could miss new combinations of work.
Limits and uncertainty
The results are descriptive comparisons, not proof that AI caused shifts in demand, budgets, or hours. OpenAI says ChatGPT usage alone does not explain posting changes and that current use does not represent the full range of AI capabilities. Upwork marketplace postings are not a census of freelance work or employment across the economy. The announcement does not establish whether the larger project budgets translate into higher freelancer earnings, stronger margins, or better outcomes for clients. Those effects, and how well the patterns generalize beyond the studied marketplace and period, remain unconfirmed.
Skill to build now
For a project involving AI, write a one-page scope that names the final deliverable, component tasks, handoffs, and review criteria before estimating cost. Compare the full bundle with a non-AI baseline, including the hours needed for editing, integration, and verification; the research’s project-level measures make a single-task estimate especially liable to miss scope.
Reader takeaway
The study does not show a simple story of AI replacing freelance work. It finds that AI-related projects tend to bundle more varied and historically uncommon tasks, alongside larger fixed budgets and more hours. Teams should price and evaluate the whole deliverable, while treating marketplace correlations as signals—not causal proof.
Open question
Do the larger budgets and hours in AI-related projects produce better client outcomes or higher freelancer earnings, once the full project scope is accounted for?