AI Didn't Kill Creativity. We Did.

The Death of Authenticity — Part 6

In 2011, a researcher published a finding that should be impossible if the popular story is true.

She'd analyzed nearly 300,000 creativity test scores from American children and adults, going back to the 1960s. Creativity had been rising for decades, right alongside IQ. Then, starting around 1990, it turned. IQ kept climbing. Creative thinking started falling, and the youngest kids fell the furthest.

That paper came out in 2011. ChatGPT didn't exist yet. It wouldn't for another eleven years.

The Claim

The story going around is simple and satisfying: AI is killing human creativity. Feeds are filling with generic content, real artists are getting drowned out, and originality is dying under a flood of machine-made sameness.

It's a clean story. It's also missing thirty years of evidence.

The Investigation

Start with music, because someone actually measured it. Researchers analyzed 464,411 popular songs recorded between 1955 and 2010, the entire pre-streaming, pre-AI era, and found a "progressive homogenization" underway: chord progressions narrowing, melodies growing more repetitive, songs getting steadily louder and more sonically similar to each other, decade over decade.

That trend was already locked in before Spotify existed, let alone before any AI could compose a track.

Fast forward to the algorithm era, and it kept going, for entirely human reasons. A 2026 comparative study of hip-hop lyrics, structure, and sound from the pre-streaming 2000s versus the streaming-dominated 2020s found songs had become measurably more generic across nearly every category researchers checked: structure, sampling, rhythm, vocal delivery, lyrical content. The researchers' explanation wasn't AI. It was economics, platforms reward "predictability" because predictable content is what keeps people scrolling, and artists increasingly write for the algorithm before they write for anyone else.

The Contradiction

Here's the problem with "AI killed creativity" as a clean, tidy story: the timeline doesn't fit.

If generative AI were the cause, the flattening should start showing up around 2022 or 2023, when AI tools got good enough to matter. Instead, the data shows creative test scores turning downward in 1990, and content homogenization accelerating steadily through the 2000s and 2010s, driven by streaming platforms, algorithmic recommendation, and audience behavior, none of which needed a machine to generate anything.

The Human Question

This is the part that's uncomfortable to sit with, because it points back at the reader, not the technology.

Every time you skip past something strange, slow, or unpolished, something a real person clearly took a real risk making, in favor of something smooth, familiar, and instantly satisfying, you're casting a small vote. Multiply that vote by a few billion people doing the same thing, a few times a day, for over a decade, and you get exactly the platforms we have now: systems built to give us more of whatever we already proved we'd reward.

The Uncomfortable Question

If a genuinely original, well-made piece of human work got buried last month by an algorithm in favor of something safe and formulaic, was that AI's doing, or the sum of a few billion ordinary people, including possibly you, quietly rewarding the familiar over and over again?

The Other Side

In fairness to a more careful version of the "creativity crisis" story: the science underneath it is genuinely contested, not settled. A 2021 re-analysis of the same type of test data, using later norms and more than 330,000 additional participants, found no evidence of an ongoing generational decline at all. Researchers still disagree about whether creativity is actually falling, whether the original test even measures the right thing, and how much of any of this connects to media, technology, or something else entirely.

That matters here. It means neither "creativity is dying" nor "AI is the reason" is as settled as either side of this argument likes to claim. What's better supported is narrower and less dramatic: algorithmic curation systems have measurably rewarded homogeneity in specific creative domains like music for over a decade, and generative AI is now accelerating output inside that same reward structure, not necessarily inventing a new one.

The Verdict

Partly true, mostly misleading.

It's misleading to say AI killed creativity, as a clean causal claim, the documented flattening in both cognitive testing and actual creative output predates generative AI by decades, and the underlying "creativity crisis" science is itself disputed. What holds up better: platforms have rewarded predictable, algorithm-friendly content for a very long time, human choices built that incentive structure, and AI tools are now the fastest engine anyone has ever plugged into it.

The Authenticity Test

Is it real? Did this come from someone genuinely exploring an idea, or was it reverse-engineered from what already performs well?

Is it human? Did a person shape this, even with AI assistance, or did a system generate it end to end, optimized purely for engagement?

Does it matter? Given honest answers to both, is this actually worth your attention, or just familiar enough to keep it?

Your Turn

The next time you scroll past something strange, unpolished, and made by a real person taking a real risk, in favor of something smooth, familiar, and safe, who do you think is doing more to kill creativity: the tool that made the smooth thing, or the thumb that kept scrolling?

Sources & Further Reading

• The Creativity Crisis: The Decrease in Creative Thinking Scores on the Torrance Tests — Creativity Research Journal (Kim, 2011)
• Something strange happened to American children after 1990 — ScienceBlog.com
• The Creativity Crisis — Newsweek
• Homogenization of pop music study (Serrà et al.) — AltRockChick summary
• From art form to asset: our study found popular songs are becoming more generic — The Conversation
• Homogenization of Pop Music: How Social Media's Algorithms Prevent Creativity and Innovation — ResearchGate