Why Do People Hate AI Art?

The Label Problem: When the Same Artwork Gets Judged Differently

Imagine two identical paintings hanging side by side. Same composition, same colors, same brushstrokes — the only difference is a small label reading "AI-made" on one of them. In study after study, that label alone is enough to change how people judge the art.

The strongest evidence comes from a six-experiment study of 2,965 participants published in Nature Scientific Reports. Over 70% of participants admitted they couldn't have told AI and human images apart without labels — yet they still rated AI-labeled works as less skillful and less expensive. The effect was largest on perceived skill and monetary value, not aesthetics. Even uncertainty hurt: "mystery" images that might be AI were rated as less skillful than unlabeled ones.

Researchers at UBC Sauder found the same pattern. As Guanzhong Du explained, "It's not like some people prefer Coke and some prefer Pepsi. It represents a deeper philosophical question about our understanding of human identity." The bias was strongest in people who believe creativity is what makes humans unique.

A peer-reviewed follow-up study confirmed the label effect on four dimensions:

• Liking — participants simply enjoyed identical works less
• Beauty — the same painting looked worse under the AI label
• Profundity — works seemed less meaningful when "AI-created"
• Worth — people were willing to pay less for them

Here's the twist that matters for business: this bias follows people into the courtroom. In an experiment with 397 participants, mock juries were more likely to find copyright infringement — and recommend lawsuits — when ads were AI-generated, awarding average damages of $7,719 versus $5,931 for human-designed ads. That's roughly 30% higher, even though style copying alone isn't legally infringement. As UGA's Mike Schuster put it, people may apply the law differently when AI is involved, "even if AI's use is legally irrelevant."

Notably, 87% of participants in the Nature study still considered most AI-labeled images to qualify as art — the objection isn't that the output looks bad. It's about what the label signals: the perceived absence of human effort, story, and emotion. The bias even proved malleable; watching robots physically produce art reduced negative judgments.

For teams building AI workflows, awareness is the practical takeaway — knowing the bias exists is the first step to judging output on merit. If you're equipping agents to work alongside you, SkillzBench is building a free, verified directory of agent skills with sandboxed benchmarks, so you can test what actually helps before you trust it.

Four Root Causes: Identity, Effort, Money, and the Definition of Art

When researchers showed people the same painting twice—once labeled "human-made," once labeled "AI"—viewers consistently preferred the human version. That result, replicated across multiple experiments, reveals that the backlash against AI art isn't really about aesthetics. It's about four distinct grievances that often get lumped together.

1. Identity threat. UBC research found the bias is strongest in people who believe creativity uniquely separates humans from other species. As UBC's Guanzhong Du put it, "It's not like some people prefer Coke and some prefer Pepsi. It represents a deeper philosophical question about our understanding of human identity."

2. Perceived absence of effort and emotion. A six-experiment study of 2,965 participants found over 70% couldn't tell AI images from human ones without labels—yet still devalued AI-labeled work, with the largest effects on perceived skill and monetary worth. Even mere uncertainty hurts: images that might be AI were rated less skillful than unlabeled ones.

3. Economic grievance. Artists object to training on copyrighted work without consent, style replication at near-zero cost, and job displacement. The stakes are concrete:

• Roughly 25 lawsuits against generative AI companies are pending in US courts
• Christie's first AI art auction drew a cancellation letter with over 6,500 signatures, per Brookings
• An experimental study found juries award roughly 30% higher damages against brands using AI-generated imagery—$7,719 vs. $5,931

4. Definitional objection. Some experts simply deny AI output qualifies as art. Architect Moshe Safdie argues AI "can imitate something that's already been created and regurgitate it in another format, but that is not an original work." Musician Yosvany Terry adds that AI compositions "lack surprise, emotion, and even silence."

Notably, the bias is malleable. Watching robots physically produce art reduced anti-AI judgments in one peer-reviewed study, suggesting perceived effort matters more than actual origin. And 87% of participants in the Nature study still considered most AI-labeled images to qualify as art. For teams working with AI rather than against it, awareness of these biases is the first step—and developers building AI workflows can explore and test free agent skills on SkillzBench's directory to make those collaborations more deliberate.

The Backlash in Numbers: Lawsuits, Petitions, and a $2.5B Market

The fight over AI art has moved from social media arguments to federal courtrooms, and the numbers tell the story. Roughly 25 lawsuits against generative AI companies are winding through US courts, filed by artists who say their copyrighted works were used to train image generators without consent or compensation.

The most consequential case is Andersen v. Stability AI, brought by ten visual artists against Stability AI, Midjourney, DeviantArt, and Runway. In August 2024, a judge allowed the infringement claims to proceed, writing that Stable Diffusion "by operation by end users creates copyright infringement and was created to facilitate that infringement by design," as Reuters reported. Whether training on copyrighted images qualifies as fair use remains unresolved — a legal gray zone every brand using AI imagery now inhabits.

The backlash extends well beyond the courtroom:

• Christie's first AI art auction prompted a letter with over 6,500 signatures demanding the sale's cancellation, even as one AI piece, "Machine Hallucinations," sold for over $270,000.
• More than 400 Hollywood insiders signed an open letter opposing copyright exceptions for AI training.
• The US Copyright Office received over 10,000 public comments on AI and copyright policy.

Meanwhile, the market keeps growing. According to Brookings analysis, the generative AI art market is projected to grow 42% through 2029, reaching over $2.5 billion — a fraction of the $65 billion contemporary art world, but expanding fast enough to keep tensions high.

For brands, the stakes are concrete and measurable. A University of Georgia experiment with 397 participants found that people were more likely to find copyright infringement and recommend lawsuits when advertisements were AI-generated, awarding average damages of $7,719 versus $5,931 for human-designed ads — roughly 30% higher. As UGA's Mike Schuster notes, "People may apply the law differently when AI is involved — even if AI's use is legally irrelevant."

The practical takeaway for business readers: AI-generated imagery carries real legal and reputational exposure, not just aesthetic risk. Companies building AI-driven creative workflows — including teams using agent skills to automate image and content pipelines — should treat provenance, licensing, and disclosure as core parts of the process. SkillzBench's free directory and MCP connector, currently in development, will let teams discover, security-scan, and benchmark the agent skills behind those workflows before deploying them against a brand's name.

Why the Bias Is Malleable — and How to Work With AI Instead

Here's the genuinely encouraging part: the research shows this bias isn't hardwired. It responds to context, effort, and framing — which means businesses and creators can work with AI visuals strategically rather than defensively.

In one peer-reviewed study, watching a robot physically produce artwork reduced anti-AI judgments, suggesting that perceived effort matters enormously to how people evaluate AI-made work (https://pmc.ncbi.nlm.nih.gov/articles/PMC10319694/). When viewers can see labor and intention behind a piece, the "AI" label loses much of its sting.

Even more striking, AI can make human work look better. In a six-experiment study of nearly 3,000 participants, the same artwork was rated more creative when labeled human-made and shown alongside AI-labeled works than when compared against other human-labeled pieces. AI comparison, in other words, can enhance perceived human creativity rather than diminish it.

UBC researcher Guanzhong Du's recommendation is simple: become aware of the bias, then collaborate with AI anyway (https://news.ubc.ca/2023/08/people-dislike-ai-art-because-it-threatens-their-humanity/). For businesses using AI-generated imagery, that awareness translates into practical choices:

• Show the human in the loop. Document the prompting, curation, and refinement behind AI visuals — perceived human engagement improves appraisals.
• Use AI where skill perception matters least, and reserve human-made work for high-stakes brand moments where the label bias carries real cost.
• Stay legally careful. With roughly 25 AI-related lawsuits pending in US courts, and juries awarding about 30% higher damages against brands using AI imagery, provenance and licensing diligence pay off (https://www.terry.uga.edu/ai-art-may-lead-to-more-lawsuits/).
• Be transparent. Over 87% of participants in one study still considered AI-labeled images to qualify as art — honesty about process beats concealment (https://www.nature.com/articles/s41598-023-45202-3).

The same awareness-first mindset applies to building AI workflows more broadly. Whether you're a creator experimenting with visual tools or a developer wiring agents into your pipeline, the skills you give those agents determine how much human judgment they amplify. SkillzBench is building a free, independent directory of agent skills — with security scanning, sandboxed benchmarks, and a free MCP connector — so you can see which skills actually lift performance before trusting them with your work. Join the waitlist to be first in line when it launches.