Freelancing and AI: What Does the Evidence Say About the Freelancer?

After ChatGPT's launch in November 2022, researchers studying the Upwork platform observed that monthly jobs for freelancers in writing and proofreading fell about 2% and their earnings 5.2%, and that in image-related work, after the appearance of DALL-E and Midjourney, jobs fell 3.7% and earnings 9.4%. The researchers stated that their study concerns only the short run, and that it was a paper under review when announced. What is the wider picture?

What was measured about demand?

Another study, published in Management Science, found that listings for automation-prone jobs, such as writing and programming, fell about 21% relative to manual jobs within eight months of ChatGPT's appearance, and that listings for image creation fell about 17%. The researchers noted that the remaining work was on average more complex and better paid, and that competition among freelancers intensified.

What was measured about performance?

In an experiment with 444 professionals on writing tasks, the tool cut completion time by about 37% and raised quality, and it narrowed the gap between participants because the weaker benefited more. In an experiment with Boston Consulting Group involving 758 consultants, those using the tool completed 12.2% more tasks and about 25.1% faster on tasks inside the "frontier" the tool handles well, but on a complex managerial task outside it they were about 19 percentage points less likely to produce a correct solution. The researchers called this the "jagged technological frontier": the tool helps here and harms there, even within a single piece of work.

The upshot is that a tool may speed up a writer and then lead them into an error they cannot see, so the freelancer's value lies in knowing where to trust and where to check.

The wider labour-market picture

The World Economic Forum's 2025 report estimates that 59 out of every 100 workers will need training by 2030, that skills in AI and big data, networks and cybersecurity, and technological literacy are the fastest growing, and that analytical thinking is the first core skill. These are estimates based on a survey of employers, not a prophecy.

It is worth noting that the studies above concern particular platforms and relatively short periods, and their results may change as the tools merge into everyday work software. They are therefore best read as indications of a direction, not final verdicts on the fate of professions.

In Saudi Arabia

In January 2025 the Ministry of Human Resources and Social Development announced that the Freelance Work Platform had registered more than 2.25 million individuals up to September 2024, and that freelance work contributed about SAR 72.5 billion to the 2023 gross domestic product, roughly 2%. The figures concern registrants, not those actually working, and say nothing of AI's effect on them. Someone who masters Arabic and understands the local context may have an edge over tools trained on other languages, but we found no measurement of this hypothesis, so we treat it as one.

What the freelancer can draw from this

What follows are analytical inferences supported in part by the evidence above, not a proven recipe:

• A narrow speciality: general work is easier for the tool; precise expertise is harder to replace.
• Verify the output: the existence of a jagged frontier makes review a skill in its own right.
• Understand the client's need before executing: execution is the part the tool shares with you.
• Build your reputation: published work and a satisfied client cannot be copied by an algorithm.
• Be honest about using the tool: tell your client what needs telling; keeping a contract is part of trust.