Freelancers have long been hailed as the agile foot soldiers of the digital economy. But when ChatGPT launched out of nowhere in late 2022, many of these independent workers found themselves adapting to a different kind of disruption: Their competition is no longer solely human.

New research from Wharton management professor Manav Raj shows how the release of the generative AI chatbot upended the online labor market by quietly and dramatically altering how workers carry out tasks, and accordingly, how they compete for jobs.

The study analyzes millions of records on a leading global freelancing platform, before and after ChatGPT’s launch. Co-authored by Wharton PhD student Shun Yiu and NYU Stern professor Robert Seamans, the research shows how the competitive dynamics on the platform evolved, with new freelancers pouring in and incumbents recalibrating how they worked to stay competitive.

The data covers about 6.6 million records showing how 312,143 freelancers applied for jobs each month.

“The key story here isn’t that jobs disappeared overnight,” said Raj. “It’s that workers shifted — some subtly, others dramatically — in response to a new kind of competition.

Gen AI Is Narrowing the Skill Gap

Freelancer behavior changed almost immediately. While total earnings and contracts on the platform remained steady, incumbent freelancers — those already active at least three months before the arrival of ChatGPT — began submitting markedly fewer job applications. Across the board, these independent contractors submitted between 51% to 62% fewer bids post-ChatGPT, a drop that persisted for nearly a year.

The reasons for this decline are nuanced, the researchers say. Freelancers in fields most directly threatened by AI, such as copywriting or language translation, have experienced a contraction in demand. These workers were far more likely to move “horizontally” into entirely new categories of work, compared with freelancers in fields that did not experience such a contraction.

But in other areas, like software development and data analytics, while demand was less affected, the platform saw a “pronounced” increase in the number of freelancers bidding for work. This increase was particularly prominent among international freelancers, who may have used new AI tools to get around barriers like language or not having formal training.

“Our findings, perhaps, are more in line with the story where AI is leveling the skill deficit.”— Manav Raj

In these cases, incumbent freelancers were more likely to reposition themselves “vertically,” bidding for proportionally fewer high-value jobs (i.e., jobs with contract values over $1,000) and more mid- to low-value jobs, as they attempted to navigate a more competitive environment.

These strategic shifts were especially noticeable among higher-skilled freelancers, as measured by the skill level required for previously completed jobs. Such workers were much more likely to shift the proportion of their bids allocated to low-to-mid-value jobs relative to higher value jobs. Rather than abandon their expertise, the researchers said many adjusted by repositioning themselves vertically to a different segment of the market.

This pattern, according to Raj, is consistent with broader evidence suggesting that AI has narrowed skill gaps.

“Our findings, perhaps, are more in line with the story where AI is leveling the skill deficit,” he said. “It’s allowing lower-skilled workers to compete with higher-skilled people, which may be why higher-skilled workers are applying for a broader range of jobs now.”

The Demand for Freelancers Remains Steady

Some online marketplaces, including Fiverr and PeoplePerHour, have started rolling out AI tools aimed at supporting freelancers through these changes. The tools either help match talent with jobs faster or automate parts of the work process. Uma, Upwork’s Mindful AI, was introduced as an intelligent work companion that helps freelancers craft proposals and manage projects, while assisting businesses in evaluating candidates and finding the right talent — streamlining the hiring and collaboration process.

“The platforms are very aware of the fact that many freelancers rely on this income,” Raj added. “There’s an understanding that we can’t hide from AI — we need to adapt.”

“I don’t think we’re at a point where freelance work is under existential threat, but the demand for it will shift.”— Manav Raj

Still, the Wharton professor believes the adjustment shouldn’t fall entirely on freelance workers or work marketplaces. “There is definitely a role for governments here,” he said. “We might think tech change is a net positive and helps us grow, but there’s this idea of creative destruction. Governments can help mitigate this friction and help people move on to other work.”

Looking ahead, he sees further disruption as likely, though not necessarily disastrous. “If AI becomes more autonomous, it will continue to affect freelance work and employment as a whole,” he said. “I don’t think we’re at a point where freelance work is under existential threat, but the demand for it will shift.”

For now, freelancers are adjusting. Some are bidding less, others are testing new roles or lowering their rates. What they are not doing — for the most part — is disappearing.