GPT-4 just beat 72% of humans on creativity tests. If you’re a mid-level creative professional, that’s not a milestone — it’s a pink slip.
A January 25, 2026 study published in Scientific Reports tested 100,000 humans against leading AI models on the Divergent Association Task, a standard measure of creative thinking. When researchers cranked GPT-4’s temperature settings to maximum — essentially telling it to take more risks — it exceeded the creativity scores of nearly three-quarters of participants. Claude and Gemini performed similarly.
The real story isn’t that AI got creative. It’s that the middle class of creativity just discovered it’s expendable.
The average creative just became expendable
This isn’t a philosophical debate about machine consciousness. It’s cold economics. The January 2026 findings add creative professionals to the growing list of high-skill jobs facing AI displacement, a trend already reshaping white-collar work across industries.
Mid-tier copywriters, junior designers, content marketers — the people who produce “good enough” work at scale — just watched their competitive advantage evaporate. GPT-4 didn’t just match average human performance on semantic distance measures. It beat most of them.
And the market knows it. Digital marketing teams are quietly restructuring around AI workflows, positioning human creatives as editors rather than generators. The 72% threshold isn’t arbitrary — it’s the line between “this tool helps me” and “this tool replaces me.”
But here’s where it gets brutal.
Elite creators just gained an unbeatable advantage
The top 10% of human creators didn’t just survive the test — they dominated it. Every AI model failed to measure up when stacked against peak human creativity, according to researchers at Université de Montréal who led the study.
The performance gap widened dramatically at the 75th and 90th percentiles. While GPT-4 could hang with the 50th percentile, it cratered against genuinely original thinkers. Co-first authors Antoine Bellemare-Pépin and François Lespinasse found that elite humans possess the skills that separate elite creators from the replaceable middle — abilities AI can’t replicate even at maximum temperature settings.
This creates a new creative aristocracy. The top 50% of humans beat all tested models on divergent thinking tasks. The top 10% weren’t just better — they operated in a different category entirely, producing semantic connections AI couldn’t touch.
Translation: If you’re genuinely exceptional, you just became more valuable. If you’re average, you’re competing with a tool that costs $20 a month.
The homogenization problem nobody wants to discuss
Here’s the catch everyone’s ignoring: AI creativity at scale creates cultural stagnation.
A January 2026 study by Arend Hintze, Frida Proschinger Åström, and Jory Schossau warns that autonomous AI generates converging outputs when deployed without human intervention. The models recycle patterns from their training data, creating a feedback loop of mediocrity. Deploy GPT-4 across a thousand marketing teams, and you get a thousand variations of the same idea.
The cultural costs of AI-generated content extend beyond individual job losses to systemic creative stagnation. When “good enough” becomes free and abundant, the economic incentive to produce genuinely original work collapses for everyone except the elite.
Despite benchmark victories, AI’s real-world performance gaps suggest the displacement timeline may be longer than the 72% statistic implies. But the direction is clear.
The test results are in. AI makes “good enough” creativity free and abundant, while elite human creativity becomes priceless and scarce. The only question left is which percentile you’re in — and whether you can prove it.
