5 min read
Here is the number that should change how you quote in 2027: 43 out of 63. That is how many freelance work categories, in a new study by OpenAI and Upwork published on November 10, showed the same pattern. The tasks people do most often inside ChatGPT lost share of Upwork job postings, while the tasks people rarely hand to ChatGPT gained share.
The headline most outlets ran with was the money. AI-related projects on Upwork carried noticeably bigger fixed-price budgets than comparable non-AI projects, while advertised hourly rates were roughly the same. AI Weekly’s summary of the report puts the budget gap at about 52 percent and the hourly gap at about 1.5 percent.
My take, stated plainly: if you are a freelancer still selling hours, this study is a polite warning that the extra money in AI-era work is going to scope, not to rate. But it is also easy to over-read, so let’s separate what the data shows from what people will claim it shows.
What the study actually measured
The report, part of OpenAI’s Work at the Frontier research series, matched two data sets: what people ask ChatGPT to do, and what clients post on Upwork. It held the type of work and the posting month constant, then looked at how the mix shifted.
Three findings matter for a solo professional:
- Demand moved away from what ChatGPT does well. In 43 of 63 categories, the activities that are common in ChatGPT lost share on Upwork. In the other 20 the pattern ran the other way, so this is a tendency, not a law.
- AI-related job posts got broader. Between early 2023 and late 2025, the typical AI-related posting went from roughly 15 equally common tasks to roughly 21, about 45 percent more varied. Clients are bundling things like video, editing and motion graphics into one brief.
- Bigger projects, same hourly rates. AI-related projects had larger fixed-price budgets and involved more hours, while posted hourly rates barely differed. The authors suggest the higher budgets partly reflect a wider scope of work.
The authors are careful to say that “ChatGPT use alone does not explain changes in freelance demand.” Keep that sentence in your pocket. It will be useful when someone on social media tells you this study proves AI is killing freelance writing.
Three things this study does not prove
It does not prove AI replaced freelancers. A category losing share of postings is not the same as a category losing work. Overall demand, platform growth and economic conditions all move the denominator.
It does not prove you can charge more per hour for “AI skills.” The opposite, if anything. Rates for AI-related work were nearly identical to non-AI work. Putting “AI-powered” in your profile headline is not a raise.
It does not describe your clients. Upwork is one marketplace with its own mix of buyers. If most of your work comes from referrals and long-term retainers, the pattern may arrive later or look different. Use the study as a signal, not a forecast for your business.
Where the money is actually moving
Once you strip out the noise, one message survives: clients are paying more for bigger, combined outcomes, and about the same for each hour of labor. That has three practical consequences.
1. The single-task gig is the most exposed
“Write a 600-word blog post.” “Transcribe this call.” “Resize these images.” These are exactly the jobs a client can now try in ChatGPT first. If that is the bulk of your work, the study suggests the pressure on your volume is real, even if your hourly rate looks stable.
2. Bundles are where budgets grow
The broader task mix is the opportunity. A client who used to hire a writer, a designer and a video editor separately is now posting one brief that wants all three, often with some automation stitched in. The freelancer who can deliver the whole bundle, using AI for the parts it handles well, captures a budget that no single-skill freelancer can.
3. Fixed price beats hourly when AI makes you faster
This is the uncomfortable math. If AI cuts a task from six hours to two and you bill hourly, you just cut your own invoice by two thirds. If you quote a fixed price for the outcome, the time you save is yours. We made a related argument in our piece on why saved hours alone will not grow a solo business.
A quoting reset for January
If the study rings true for your market, here is how I would rework a freelance offer before the new year:
- List your last ten jobs. Mark each one as “single task” or “outcome”. Be honest about which ones a client could now attempt alone with a chatbot.
- Design one bundle. Combine your core skill with one or two adjacent ones you can deliver with AI help. A copywriter might sell “launch kit: landing page, five emails, and three short video scripts.” A bookkeeper might sell “year-end close plus a cash flow forecast.”
- Price the outcome, not the hours. Anchor on the value to the client and what they would pay three separate freelancers, then come in under that total.
- Keep hourly for open-ended work only. Advisory calls and messy, undefined projects still suit hourly billing. Everything repeatable should move to fixed scope.
- Decide how you handle AI costs. Bundled pricing makes this simpler, because tool costs are absorbed into the package. Our guide on whether to bill clients for AI costs walks through the options.
The skeptic’s bottom line
This is one study, from a company with an obvious interest in showing that its product creates rather than destroys work, using data from one marketplace. Its own authors call the results mixed. That is precisely why the narrow conclusion is credible: it does not claim much. It simply shows that broader scopes are getting bigger budgets while the price of an hour stays flat.
You do not need to believe any grand story about the future of work to act on that. You only need to stop selling the unit that is not getting more valuable. If you want help putting new numbers on your offer, our 90-minute workflow for raising prices in January is a good place to start.
Are your clients already asking for bundled work? Share what you are seeing in the comments, especially if it contradicts the study.
