AI & Creativity: A Complex Debate

AI & human creativity

The evidence points to a genuine tension

Unlike the rest of this index, the creativity question doesn't resolve into "problem" or "non-problem." The evidence points to a genuine tension: AI can enhance individual creativity while, at scale, reducing collective diversity. Both can be true at once.

56% peak adoption, startups 2 credible studies 4 credibility criteria
Enhances individual creativity

Multiple studies show AI tools help individuals generate more ideas, improve quality, and cut the time to complete creative work.

Risks collective homogenization

The same assistance, adopted widely, can push outputs toward similarity — lowering the diversity of the creative ecosystem overall.

AI-tool adoption varies sharply by creator type

Share using AI tools weekly or daily, by organization type. *35% of established brands report never using AI — the inverse of the startups/agencies/freelancers who've largely adopted it.

Startups
56%
Agencies
53%
Freelancers
51%
Established brands*
35%
Startups · Agencies · Freelancers (largely adopted) Established brands* (the inverse)

Source: LinkedIn & Filestage, 2024 industry survey; adoption & approaches also per Belsky, 2023 (Adobe) and Doshi & Hauser, 2024

What the strongest studies find

StudyBelsky, 2023 · Adobe
Context dependent

Identified two creator approaches to AI: outcome-oriented individuals focused on final products, and process-oriented creators using AI within a broader creative practice.

StudyDoshi & Hauser, 2024
Both effects

The pivotal finding: the same study provides strong evidence for both enhanced individual creativity and increased homogenization at the collective level.

The core insight — a "social dilemma"

What benefits an individual creator may, if widely adopted, produce a less diverse creative ecosystem. The debate isn't binary; it's a tension between individual and collective outcomes.

Why these studies are credible

  • Experimental designs with control groups
  • Pre-registered methods (reducing publication bias)
  • Field experiments in real-world settings
  • Mixed-methods and standardized measurements across conditions

Part of the AI Problems Index · see the Risk Atlas and Environmental Impact.