in ,

AI Freelancing Pays in Bigger Projects, Not Higher Hourly Rates: What the OpenAI and Upwork Study Found

AI Freelancing Pays in Bigger Projects, Not Higher Hourly Rates

52 percent. That is how much larger fixed-price budgets were for AI-related Upwork jobs than for comparable non-AI jobs, once researchers from OpenAI and Upwork controlled for job category and posting month. After the same controls, the gap in posted hourly rates was about 1.5 percent. AI freelancing pays more because the projects are bigger and combine more tasks, not because clients pay a higher price per hour, so freelancers who want the upside should price by scope and outcome rather than by rate.

Those figures come from The Rebundling of Freelance Work in the AI Age, a November 2026 paper by Caroline Chin, Alex Martin Richmond and Carl Shan of OpenAI and Shagun Kala, Nadav Kunievsky and Teng Liu of Upwork. OpenAI published a summary as part of its Work at the Frontier series. Because it pairs ChatGPT usage with real job postings, it says more about what clients actually buy than any survey of freelancer opinions can.

Hosting 75% off

What the researchers measured

The team classified about 36,900 sampled consumer ChatGPT conversations from January 2025 to June 2026 into Upwork’s 64 work subcategories, then compared them with Upwork job postings through 2025 and the first half of 2026. The task-level analysis uses roughly 104,000 postings, a 5% sample, broken into more than 400 standardised tasks. Upwork’s own classifier decided which posts count as AI-related.

The authors are careful about limits. The analysis is descriptive, not causal, the two datasets are not matched person to person, and the top 1% of posted prices were trimmed. They also do not break results down by country, so nothing here is specific to Pakistan. Read the numbers as a map of where client money is moving, not a promise about any single profile.

Budget up, rate flat: three numbers to remember

The price findings split cleanly into three parts:

  • Fixed-price budgets: about 70% higher for AI-related jobs before adjustment, and about 52% higher after controlling for subcategory and month.
  • Posted hourly rates: about 9% higher before adjustment, falling to roughly 1.5% after controls.
  • Hours worked: on hired hourly contracts, AI-related work logged about 41% more hours in the first 90 days, again after controls.

Put together, clients are commissioning more work per project rather than paying a premium per hour for “AI” on a profile. The authors compare this to the Jevons rebound effect: when AI makes a kind of work cheaper to do, buyers ask for more of it rather than simply spending less.

For a Pakistani freelancer who has spent years being undercut on hourly rate, this is useful news. If the rate premium for AI work is close to zero, competing on hourly price was never the lever. The lever is the size of the job you can credibly own.

Why are AI jobs bigger?

The paper’s answer is in its title: rebundling. AI changes the cost of doing, coordinating and handing context between tasks, so clients start buying combinations that used to go to separate people. Several findings point the same way:

  • AI-related posts list an average of 3.35 distinct tasks against 3.19 for other posts, and are more likely to contain three or four tasks rather than one or two.
  • Historically uncommon task pairs make up 8.4% of the pairs inside AI-related jobs, compared with 2.4% in other jobs. The gap is widest in generative AI and creative production.
  • The effective number of task clusters in AI-related postings rose from 14.6 to 21.2 between early 2023 and late 2025, about 45% more diverse.
  • Within AI postings, video production, social media campaign management, marketing automation and AI API integration gained ground, while “build chatbots and machine-learning models” slipped from 21.8% to 17.8% of the mix even though it stayed the largest cluster.

The authors conclude that value increasingly sits in defining the problem, connecting tools, applying judgment and delivering a finished outcome. A client who once hired a scriptwriter, an editor and a social media manager separately may now want one person who produces the video, cuts the short versions and schedules the campaign.

Where demand moved on the marketplace

Comparing the first half of 2025 with the first half of 2026, the categories with the largest share gains were Mobile Development (up 2.27 percentage points), Video and Animation (up 1.38) and QA Testing (up 0.58). The largest declines were Web Development (down 1.98), Web and Mobile Design (down 0.68) and Content Writing (down 0.57). Content writing is the starkest mismatch in the study: it makes up about 14.5% of the ChatGPT conversations sampled but only about 1.3% of Upwork postings, and its postings fell at more than twice the rate of the overall market over two years.

Upwork’s own third-quarter 2026 results show the same split at company level. Total gross services volume fell 10.1% year on year to $914.1 million, while work by AI-skilled freelancers reached an estimated annualised run rate of nearly $450 million, up 55%. If you want the skill-by-skill view, our breakdown of Upwork’s 2026 skills data covers which AI skills are growing fastest.

Reprice your offer around the bundle

Turning the research into a pricing change takes four moves:

  1. Name the outcome, then list the tasks inside it. “Customer support automation” reads as one purchase. Underneath it you might deliver the chatbot, the help-centre articles it draws from, the handoff to WhatsApp for edge cases and two weeks of testing. Quote one fixed price for the set, written as $[your figure], rather than an hourly number for each piece.
  2. Add one adjacent task you can do well. The study shows clients pairing tasks that rarely appeared together before AI. If you edit video, adding captioned short cuts and a posting schedule turns a $[X] edit into a campaign. If you build apps, adding AI API integration and test cases mirrors the categories that gained share.
  3. Use hourly contracts for ongoing scope, not for discovery. The 41% extra hours finding suggests AI projects grow after they start. A fixed-price first milestone followed by an hourly or retainer phase captures that growth without arguing about your rate.
  4. Check your net, not your gross. Bigger contracts matter only if the margin survives platform fees, and Upwork’s take rate is worth recalculating against every bundled quote.

The group this study should worry most is freelancers selling a single, easily described task that also happens to be a common ChatGPT use, with general content writing as the clearest case. The way out is to move that skill inside a bundle where the client pays for the result, such as writing plus video scripting plus distribution, or writing plus SEO plus marketing automation, rather than to cut the rate.

The paper calls itself an early baseline for how more agentic AI may change freelance work. That framing fits a practical next step: send bundled, fixed-price offers on your next five proposals and compare the budgets you win against the hourly jobs you used to chase.

Hosting 75% off
Upwork Now Explained: How Pakistani Freelancers Can Win Instant Hires

Upwork Now Explained: How Pakistani Freelancers Can Win Instant Hires

DigiSkills Batch-05 Courses: Which Two to Choose for Freelancing

DigiSkills Batch-05 Courses: Which Two to Choose for Freelancing