Scientists Just Tested AI Against 100,000 Humans on Creativity. AI Won. Here Is What That Actually Means — and What It Doesn't.

In January 2026, ScienceDaily published a study that immediately generated headlines no AI researcher wanted to write but could not avoid: 'Researchers Tested AI Against 100,000 Humans on Creativity. Generative AI Can Now Beat the Average Human on Certain Creativity Tests.' The study, conducted by researchers at a major US university and published in a peer-reviewed journal, is the largest comparative assessment of AI and human creativity ever conducted. Its findings are real, significant, and consistently being misunderstood in both directions — by people who are using it to declare human creativity obsolete and by people who are dismissing it as irrelevant. The truth is more nuanced and more important than either camp is acknowledging.

What the Creativity Tests Actually Measured

To understand what the study found, you need to understand what the creativity tests it used actually measure. The study used two primary assessments: the Alternative Uses Task (AUT) and the Divergent Association Task (DAT). These are well-validated psychometric instruments, but they measure specific dimensions of creativity — not creativity in its totality.

• Alternative Uses Task: participants are asked to generate as many unusual uses as possible for a common object (a brick, a paperclip, a shoe). Performance is scored on fluency (number of ideas), originality (how unusual the ideas are relative to other participants), and elaboration (how developed each idea is). AI scored higher on originality and fluency than the average human.
• Divergent Association Task: participants generate a list of words that are as semantically distant from each other as possible. Higher scores indicate greater ability to make distant conceptual connections. AI again scored above average human performance.
• What these tests do not measure: they do not measure the ability to produce a finished creative work. They do not measure taste, curation, emotional resonance, intentionality, or the judgment to know which idea from a long list is actually the one worth pursuing. These latter qualities — the ones that separate a good first draft from a masterwork — are not what these tests assess.

What the Researchers Actually Said — Before the Headlines Simplified It

The study's lead researchers were careful in their conclusions in ways that the subsequent coverage was not. Several important caveats appeared in the paper that were absent from most news coverage.

• AI performed variably across the tests: AI scored highest on tasks requiring many divergent ideas quickly and lowest on tasks requiring sustained, coherent creative development over longer contexts. The 'AI beats humans' finding applies specifically to generative fluency in brief creative tasks — not to sustained creative development.
• The population sample matters: AI outperformed the average human participant. It did not outperform highly creative humans — professional artists, experienced writers, trained musicians. The study specifically found that the top 1% of human performers on these creativity tests still outperformed AI. The finding is that AI exceeds the median, not the ceiling.
• Creativity and creative work are not the same thing: scoring high on a creativity assessment and producing creative work that moves, inspires, or has lasting cultural value are related but not identical skills. The history of art is full of technically accomplished work that has no lasting impact — and technically imperfect work (early punk, folk music, much photography) that has profound cultural resonance.

What This Actually Means for Creative Professionals

The creativity study, properly understood, changes the practical landscape for creative professionals in a specific and limited way. It confirms what many creative professionals already know from experience: AI is a powerful idea-generation tool. Given any creative prompt, AI can produce a large volume of diverse, often surprising outputs quickly — outputs that function well as raw material, starting points, or contrast cases to push human thinking in new directions.

• The implication for writers: AI generates divergent textual ideas well. The writer's job is increasingly not to generate the first idea but to recognize the right idea from among many — to exercise taste, judgment, and intention. These are human faculties that the creativity tests do not measure.
• The implication for visual artists: AI image generation has demonstrated extraordinary fluency in producing visual ideas. The artist's enduring value is the authorial vision that determines what is worth making, the technical and aesthetic judgments about execution, and the cultural context that makes a work meaningful rather than merely novel.
• The implication for musicians: AI can generate melodic fragments, harmonic progressions, and rhythm patterns with high fluency. The musician's irreplaceable contribution is the emotional intelligence, performance interpretation, and cultural meaning-making that turns notes into music that matters to humans.
• The honest bottom line: the creativity study confirms AI as a powerful creative collaborator and ideation tool. It does not confirm AI as a replacement for human creative judgment, taste, or meaning-making. The professionals who treat AI as a powerful first-pass generator and apply their own discernment to the output are more effective than those who either resist AI or outsource their judgment to it.