Since the 2022 release of ChatGPT, freelance graphic designer Lisa, based in Spain, has witnessed a profound shift in her professional landscape. Once focused on creating original logos and packaging, she now finds that 90% of her incoming requests involve correcting AI-generated content. This remediation work has become a significant pillar of her career, accounting for 60% to 70% of her annual income. Lisa, who prefers to keep her full identity private to avoid unwanted solicitations, describes the AI output as “soulless” and has shifted her personal focus toward “human-made designs.”
Lisa is part of a broader trend where creative professionals are increasingly hired to manage “AI slop”—a term for low-quality, automated output. As companies integrate tools like OpenAI’s ChatGPT and Anthropic’s Claude, they often underestimate the technical requirements of refining these drafts. While businesses may hope for quick, inexpensive fixes, freelancers report that polishing AI output for commercial use is often as labor-intensive as starting from scratch.
Market data confirms this surge in demand for human intervention. On the platform Freelancer.com, which serves over 88 million users, job listings specifically tagged with terms like “correct AI,” “AI hallucination,” and “AI error” climbed 87% to 10,760 globally between August 2025 and June 2026. Similarly, Upwork has seen a 70% year-over-year increase in AI remediation gigs, while Fiverr reported that keyword searches for “AI cleanup” services grew more than 20 times between 2023 and 2026.
This shift comes amid a difficult climate for the freelance labor market. A 2025 study in the journal Management Science found that listings for freelance roles exposed to automation dropped 21% in the eight months following the launch of ChatGPT, with image-creation gigs falling 17% after the arrival of AI image generators. These losses have persisted into 2026. Meanwhile, competition has intensified as companies reduce staff; the average project on Freelancer.com attracted 54 bids in 2025, an 8% increase from the previous year.
Matt Barrie, CEO of Freelancer.com, notes that small companies and entrepreneurs are the primary drivers of this cleanup demand. They often use AI to generate a first draft but lack the expertise to resolve the resulting errors. “[AI output] needs a human to turn it into something polished that can actually be used,” Barrie said, noting that the time saved on initial drafting is frequently consumed by the “incredibly time-consuming” process of making the work commercially viable. He added that while he finds the cleanup work “extremely distasteful,” it remains a necessary service to fix the problems AI creates.
The experience of Florida-based illustrator Todd Van Linda highlights the financial necessity of this work. He began accepting AI cleanup gigs in 2023 as his traditional commissions for video game art and architectural plans slowed. While he maintains his standard rates, he faces clients who expect the work to be cheap. In one instance, a client offered only $500 to repair 13 to 15 AI-generated illustrations for a children’s book. Van Linda noted that some clients even attempt to pass off AI images as original work, sometimes lying about their methods.
For Australian writer and editor Kym Dunbar, AI cleanup now represents 60% of her workload as traditional writing assignments dwindle. She began the work in 2024, learning to address common AI issues like repetitive language and awkward structure. While she finds the process “interesting” and “challenging,” she worries that society is beginning to view AI as the definitive end-point of creativity. “I just feel it’s a shame we need to do it,” she remarked.
The technical challenges can be immense. Multimedia editor Nathan McConnell, who takes on freelance video editing alongside his full-time job, described fixing AI-generated tarot deck images that were riddled with anatomical errors like extra fingers and two left feet. He often spends two to three hours per image in Photoshop to make them usable. Some projects, however, are beyond repair; McConnell has told clients that certain files were “too” slop to be salvaged, particularly when physics or lighting were fundamentally broken.
The future of this niche remains a subject of debate. Dunbar believes the need for human intervention may vanish within five to 10 years as AI models improve. Conversely, McConnell argues that AI will require constant “babysitting” to catch subtle errors that untrained users miss. Meanwhile, some professionals are moving away from the trend; Van Linda stopped taking cleanup gigs late last year, finding that a segment of his client base is increasingly dissatisfied with AI and is actively seeking original, hand-drawn work again. The report also notes that while others enjoy the challenge, some find it tedious and creatively unfulfilling. The report also notes that according to a 2026 study, and those losses have continued. The report also notes that improving resolution, and hunting for glitches like an extra toe, the work typically involves filtering out visual noise. The report also notes that he said, his rate is $65 an hour and “doesn’t go down from there”. The report also notes that as a result, Dunbar said her monthly earnings from writing original content have fallen to roughly a third of what she made during the pandemic. The report also notes that from repetitive language and awkward paragraph structures to a lack of emotional nuance, she quickly learned to spot recurring problems. The report also notes that since then, she has worked on about 40 AI-generated nonfiction books, longform articles and technical papers. The report also notes that “There’s a lot less good projects available for writers and editors these days,” Dunbar said.
