The same technology that was initially predicted to render graphic designer Lisa Carstens jobless has instead made her busier than ever. Carstens, a seasoned freelancer based in Spain, dedicates a significant part of her day to collaborating with startups and individual clients who seek assistance in rectifying their poorly executed AI-generated logos. The designs that clients present to her often suffer from unclean lines and nonsensical text, resembling a chaotic mess of pixels, especially when enlarged.
“There are people who understand that AI isn’t perfect, and then there are those who approach you in frustration because they couldn’t achieve the desired results themselves using AI,” Carstens explains. “You have to approach these situations with empathy; you don’t want them to feel foolish. Then, you have to get to work fixing it.” These assignments are part of a new job category that has emerged from the generative AI boom, which initially posed a threat to creative professions. While anyone can now create blog posts, design graphics, or code applications using simple text prompts, AI-generated content rarely meets the quality standards required for satisfactory final products.
This trend has significantly altered the job market for many gig workers. Despite ongoing worries that AI is replacing jobs across various sectors, numerous professionals report discovering new opportunities because of AI’s limitations. For instance, writers are increasingly hired to enhance the output of tools like ChatGPT, while artists are called upon to refine poorly generated AI images. Even software developers are finding themselves tasked with fixing buggy applications produced by AI assistants.
A recent report from MIT revealed that AI has displaced outsourced workers more than permanent employees. However, it also indicated that an alarming 95% of businesses’ generative AI initiatives yield no return on investment. “The core barrier to scaling is not infrastructure, regulation, or talent,” the report states. “It is learning. Most GenAI systems do not retain feedback, adapt to context, or improve over time.”
For Carstens, the AI-generated logos that clients send her can vary in quality. Some designs require only minor adjustments, while others necessitate a complete redesign from scratch, all while maintaining fidelity to the original AI concept. This process can be more time-consuming than creating a logo from the ground up. Many freelancers express that fixing AI’s mistakes is not their preferred line of work, as it typically offers lower pay compared to traditional projects in their expertise. Still, for some, it’s a necessary source of income.
Freelance writer Kiesha Richardson, based in Georgia, shares her experience, indicating that half of her current assignments involve clients hiring her to refine or rewrite AI-generated articles that “don’t look remotely human at all.” Some of the issues are straightforward to correct, such as AI’s repetitive use of em dashes or its tendency to favor certain words like “embark” and “deep dive.” However, beyond these quirks, Richardson points out that AI often produces generic responses that fail to provide thorough answers, necessitating additional research on her part.
Richardson notes that many clients undervalue the effort required to enhance poorly written AI articles, often offering lower pay based on the assumption that these tasks are less demanding. “Making AI sound more human can require just as much thought and creativity as writing the entire article myself,” she emphasizes. “I worry that companies are using AI to reduce costs, which could impact my pay. However, they soon realize they can’t fully rely on AI, as it doesn’t provide the quality content they expect. Hopefully, that ensures we remain necessary players in the industry.”
As companies navigate their AI strategies, recent data from freelance job platforms like Upwork, Freelancer, and Fiverr indicates a surge in demand for various creative services this year. Clients are increasingly seeking human professionals who can collaborate with AI technologies instead of fully depending on or dismissing them. Upwork’s data shows that while AI automates lower-skilled tasks, there is a rising demand for more complex work, including content strategy and creative direction.
Fiverr has reported a staggering 250% increase in demand for niche tasks in web design and book illustration. Similarly, Freelancer has observed a notable rise in requests for human involvement in writing, branding, design, and video production, including emotionally resonant content like “heartfelt speeches.” As Matt Barrie, CEO of Freelancer, remarks, “The quickest way to get rejected is to use AI to write a love letter.” This sentiment holds true for brands as well, as consumers can often detect when content is entirely AI-generated, leading to negative reactions.
Brands caught using AI without sufficient human involvement have faced backlash; for example, Guess faced criticism online after featuring an AI-generated model in a Vogue advertisement. This backlash emphasizes that, beyond the evident flaws of AI tools, many clients prefer a human touch to differentiate themselves from the growing pool of AI-generated content.
For Todd Van Linda, an illustrator and comic artist in Florida, the differences between AI-generated and human-created art are stark. He can quickly identify AI art not only by certain inconsistencies but also by the distinct plasticine effect that characterizes AI images across various styles. Independent authors often turn to him because they understand that AI-generated art fails to capture the unique essence of their stories. Van Linda’s clients typically provide only a rough idea of what they seek, making it his responsibility to interpret their preferences and create art that evokes the desired emotion.
While he occasionally receives requests to “fix” AI-generated artwork, Van Linda has opted to decline these jobs, recognizing that such clients often undervalue his work. “Fixing those images usually requires more effort than starting fresh, as they often consist of disjointed elements that don’t align with the intended vision,” he explains. “Clients are reluctant to invest more money, even when they receive subpar results.”
The financial constraints faced by clients who have already opted for cheaper AI solutions have impacted gig workers across various fields, including technical roles like coding. For Harsh Kumar, a web and app developer based in India, many clients have already exhausted their budgets on “vibe coding” tools that ultimately failed to deliver satisfactory results. However, some clients are beginning to understand that hiring a human developer can save them from the hassles associated with trying to rectify AI-generated code.
Kumar has worked on projects that involved correcting AI-powered support chatbots that provided inaccurate information and compromised security by leaking sensitive details. He has also rebuilt an AI content recommendation system that frequently crashed and delivered irrelevant suggestions. “While AI can enhance productivity, it cannot fully replace humans,” Kumar asserts. “I believe that human input will remain essential for long-term projects. After all, humans are the ones who created AI in the first place.”
