Will AI Ever Be More Creative Than Humans? A Field Guide for Creatives
A new study suggests AI's creativity has a ceiling. Using the Standard Definition of Creativity, the research found that large language models can mimic creative behavior but tend to cap out around the average human. Under current design principles, they don't reach expert-level output.
But creativity isn't a single metric. It's intent, risk, novelty, usefulness, and acceptance by an audience. That's why the debate won't end anytime soon - and why your workflow matters more than the headline.
What the research actually says
The study, published Nov. 11 in the Journal of Creative Behavior, evaluated outputs from popular models and judged them against the "novel and useful" standard.
Conclusion: "AI can mimic creative behavior - quite convincingly at times - but its capacity is capped at the level of an average human and can't reach professional standards under current design principles." As one point put bluntly, "Generating something is not the same as being creative."
Why many creatives still feel the gap
Jack Shaw, who benchmarks LLMs for marketing, draws a hard line: if creativity means reframing a brief, setting new cultural cues, and taking responsibility for risky choices, humans lead. Models optimize for likelihood; they don't carry intent or stakes.
Alesha Brown adds the missing piece: lived experience. "No LLM wakes up with a childhood trauma, a cultural lineage, or a moral conflict and decides, 'I'm going to make a film or write a book that could cost me relationships but might free other people.'" That "why" behind the work still belongs to us.
Where AI already looks creative
In applied domains, usefulness is measurable - and that changes the game. Gor Gasparyan reports AI surfaces novel keyword and theme connections for his team about 80% of the time, driving fresh content strategies they hadn't considered.
Iliya Rybchin argues both humans and models work the same way: recombining stored patterns under constraints. "We romanticize ex-nihilo creation. In reality, creativity is almost exclusively combinatorics." If creativity = connecting dots, "the entity with the most dots wins."
James Lei frames it cleanly: "Creativity is generation plus selection against a purpose." AI shines when you can generate many options and select based on clear criteria - ad concepts, onboarding flows, contract clauses, musical motifs. Gatekeepers decide what sticks.
You get out what you put in
Where AI struggles: open-ended, long-horizon agenda setting that needs lived context and cross-domain judgment. Where it works: clear instructions, defined success metrics, and tight feedback loops.
Amit Raj sees it daily: vague prompts yield bland ideas; context, challenge, and iteration pull better creative work out of the model. The system is plastic - your process shapes the output.
A practical playbook for creatives
- Set a purpose: define the audience, job-to-be-done, and the constraint (tone, length, medium, budget). Creativity needs a target.
- Write a rubric: 3-5 criteria you'll score every idea on (novelty, usefulness, clarity, brand fit, risk). Keep it objective.
- Force divergence: ask for 25-50 options across different frames (metaphor, angle, format, cultural reference). Quantity first.
- Constrain formats: use repeatable templates (headline formulas, storyboard beats, hook/turn/payoff). Constraints raise the floor.
- Stage the prompts: brief → generate → critique → revise → combine. Make the model debate itself before you step in.
- Mix data with taste: run quick tests (polls, CTR sprints, small A/Bs) to down-select, then use creative judgment on the finalists.
- Inject lived experience: add your stories, sensory detail, and stakes. Ask the model to adapt drafts to your anecdotes.
- Push for edges: request "weird but workable" options, then ask for a safer pass that keeps one surprising element.
- Build a reference brain: feed style guides, past winners, customer quotes, and swipe files. Better inputs, better recombinations.
- Respect ethics: check references, sourcing, licensing, and likeness rights. Run an originality check before publishing.
- Ship, then iterate: treat each release as a test. Archive what wins, and retrain your prompt stack on the outcomes.
What to use AI for today
- Copy and content: idea lists, headline variants, structure outlines, objection handling, value propositions.
- Design and art: concept boards, style mashups, negative space explorations, layout variations for constraints.
- Audio and video: hook scripts, beat sheets, motif exploration, alt-takes for pacing and tone.
- Operations: naming sprints, mood-board tags, brief reframes, stakeholder translations.
Mindsets that keep you ahead
- Be the director, not the camera. Your edge is taste, intent, and the risk you're willing to take.
- Think "generation + selection." The model generates; you set the purpose and select with judgment.
- Bias to shipping. Creativity compounds through feedback, not theory.
So, will AI surpass top creatives?
Depends on the definition. Under current systems, models imitate well and deliver in domains where success is measurable - but they don't carry a "why," take social risk, or set new cultural cues.
For working creatives, the move is simple: use AI to widen the option space, then apply your taste, experience, and courage to choose and refine. That combo wins briefs today - and keeps you valuable tomorrow.
Helpful links
Your membership also unlocks:
