Human creativity not at risk from AI artists
Published: 24 November 2025
A new study says there's a "mathematical ceiling" on how creative large language models can be. Translation: AI won't replace top-tier playwrights, authors, songwriters, artists, or screenwriters.
According to the research, current LLMs top out at roughly 0.25 on a creativity scale of 0 to 1. They can mimic style and produce convincing work, but they plateau at average-human originality and won't hit expert standards under their current design.
Generation isn't the same as creativity
AI generates what's expected based on patterns in its training data. That's why it's fluent and fast-and why it leans toward safe, familiar outputs.
Creativity demands new, original, and effective ideas. If the output is predictable, it's not truly creative-no matter how polished it looks.
The 0.25 cap, in plain English
LLMs predict the next word with probability math. That math rewards what's most likely, which drags results toward the mean.
Novelty lives in the tails-the unusual, risky choices. Systems tuned for likelihood naturally avoid the risky edge where breakthrough work happens.
Curious how these systems work at a high level? See this explainer on large language models for context: What is an LLM?
What this means for your craft
- Your taste, point of view, and risk tolerance are the moat. That's where the model can't follow.
- AI is useful for breadth: options, riffs, comps, structure, cleanup. You bring the spark and the standard.
- Treat AI as a studio assistant. You're still the creative director.
Use AI here (and don't here)
- Great for: rough drafts, alt lines, scene beats, references, mood boards, variations, grammar, timing, summaries.
- Use with care for: voice, theme, concept, character, core melody, visual language. That's your signature.
- Avoid outsourcing: the final concept, the core hook, or the risky choice that makes the work memorable.
A simple workflow that works
- Define intent and constraints: audience, outcome, tone, constraints you'll break.
- Generate quantity with AI. Aim for 30-50 options, fast.
- Curate hard. Merge two weak ideas, subvert the obvious, change form or medium.
- Iterate with deliberate rule-breaking. Push past the model's safe zone.
- Test with real people. Keep what moves them; cut what only reads as "smart."
Talking to clients about AI
Be clear: AI boosts speed and volume. Your value is judgment, context, and taste-the part that makes work land.
Price for outcomes and creative direction, not just deliverables. Sell the process, not the prompt.
The bigger picture
This study cuts through the hype: LLMs can simulate creativity, but they're capped by the math that makes them fluent. Useful tool, yes. Replacement for top creatives, no.
If you want to get sharper at using AI as a collaborator-without losing your edge-browse role-specific training here: AI courses by job. For day-to-day craft support, these toolkits help with fast drafts and concept sprints: AI tools for copywriting.
Source
The study references a "mathematical ceiling" on LLM creativity and reports a maximum creativity score of 0.25. Full title: "The Cat Sat on the …? Why Generative AI Has Limited Creativity."
Your membership also unlocks:
