AI freelance projects carry 52.4% larger budgets, OpenAI-Upwork study finds

AI-related fixed-price freelance postings on Upwork carried budgets 52.4% higher than comparable non-AI postings, while the adjusted gap in advertised hourly rates was just 1.5%. That contrast sits at the center of a joint study by OpenAI and Upwork, published November 10, 2026, titled The Rebundling of Freelance Work in the AI Age.

The report treats the mismatch as evidence that project scope and the quantity of labor purchased matter more than any simple hourly premium. Among hired hourly contracts with positive recorded hours, AI-related work involved 40.6% more hours during the first 90 days after hire, with observations running through April 2026.

The researchers classify a posting as AI-related when it requests AI skills, requires AI tools or involves improving AI-generated output. A fixed-price budget covers an entire deliverable, while the hourly measure uses the midpoint of the client's advertised range — separating the price of time from the amount of work purchased. All three adjusted estimates compare work within the same subcategory and posting month.

Does direct AI use substitute for freelance hiring?

Content writing offers the clearest contrast. It accounted for 14.5% of sampled personal and work-related ChatGPT conversations from January 2025 through June 2026, against 1.3% of Upwork postings in the first half of 2026. Its share of Upwork postings fell roughly 31% between the first halves of 2025 and 2026 — the equivalent of losing nearly a third of its platform footprint in one year.

Other categories tell a different story. Video and animation represented 1.0% of sampled ChatGPT conversations but 10.0% of Upwork postings; mobile development accounted for 0.2% and 7.8%, respectively.

The pattern across 63 matched subcategories resists a simple substitution narrative:

• 43 categories either had lower relative ChatGPT intensity and gained marketplace share, or higher intensity and lost share — the direction substitution would predict.
• Six combined higher intensity with share gains.
• Fourteen combined lower intensity with losses.
• Twenty fell outside the simpler pattern entirely.

Direct AI use alone, the authors conclude, does not explain the changing composition of freelance demand.

How are AI projects changing shape?

The effective number of task clusters in AI-related postings rose from 14.6 in January–June 2023 to 21.2 in July–December 2025, an increase of roughly 45%. The measure expresses diversity as the equivalent number of equally represented clusters.

Within individual postings, AI-related jobs contained an average of 3.35 distinct task clusters, versus 3.19 in other jobs. AI-related postings were more likely to contain three or four tasks, with no corresponding rise in the five-or-more category.

Unusual task pairings became markedly more common. Among jobs with at least two tasks, historically uncommon pairs — defined as pairs appearing together in fewer than 200 pre-AI postings — accounted for an average 8.4% of pairs in AI-related postings and 2.4% in non-AI postings, more than three times the rate. The report cites examples such as AI-generated video combined with editing and motion graphics, or chatbot development paired with application integration.

Building chatbots and machine-learning models remained the largest task cluster, though its share of task assignments declined from 21.8% to 17.8%. Social-media campaign management, AI-generated video, marketing automation and AI API integration gained share within AI-related work from 2023 to 2025.

What does the data actually measure?

The ChatGPT comparison drew on 36,896 conversations sampled from consumer accounts and classified as personal or work-related. It excluded schoolwork, unclear-purpose conversations and those without an Upwork taxonomy match. The model GPT 5.6 Luna classified each conversation using up to its first five available user messages; no researcher read user messages or examined individual user data.

On the Upwork side, more than 3,000 activity descriptions were grouped into more than 400 standardized task clusters, with a GPT-powered pipeline evaluating each posting's title, description, requested skills and taxonomy context.

The authors flag clear limits. The two datasets are not directly matched, neither represents the labor market as a whole, and the estimates describe associations rather than causal effects. The category analysis tracks share changes through the second half of 2025 and does not measure aggregate marketplace growth.

Separate market context comes from Upwork's July 14, 2026 Future Workforce Index, which drew partly on a March–April 2026 survey of 2,400 U.S.-based skilled workers. It found 38% of skilled U.S. knowledge workers freelance, up from 28% the prior year — a ten-point jump. That survey is distinct from the joint report and does not supply its platform estimates.

The authors' broader reading: freelance value may increasingly lie in connecting tools, applying domain judgment and delivering finished outcomes. For employers weighing contingent labor, project scope and integration now carry as much of the AI labor-market signal as demand for any single activity.