Does AI make us more creative or less? The best evidence says both. Leaders need to manage the tension on purpose.
Better stories, more alike
In 2024, Doshi and Hauser published an experiment in Science Advances. 300 people wrote short stories, some with AI ideas and some without. 600 judges rated the results.
AI ideas made stories more creative and better written. The least creative writers gained most.
But stories written with AI were more similar to each other. Each writer felt more original. The pool as a whole became more uniform.
The 2026 evidence
The pattern has held. Ju and Aral's 2026 experiment involved 2,234 people producing more than 11,000 adverts. Human-AI teams produced 50% more ads per worker, with better text.
Their output was also more homogeneous. The more work people handed to AI, the more alike the results became.
A 2026 review in Trends in Cognitive Sciences by Sourati, Ziabari and Dehghani goes further. It warns that large language models can homogenise how people write and even how they think.
Build this capability in your organisation
Growth Performance designs bespoke programmes that turn insight into measurable behaviour change.
Why sameness is a business risk
When many people draw on the same models, their starting points converge. The result is fluent, competent and familiar.
For an individual, that trade may be worth it. For an organisation that competes on being different, a slow drift to sameness is a serious risk.
Originality as a deliberate act
Avoiding AI is the wrong lesson. Weaker performers benefit most, and that is valuable.
The right lesson is that originality now needs a process. Use AI to escape the blank page. Then push deliberately past the obvious answer.
Building that into how work is done is part of Adaptation: changing the work, as well as adding a tool to it.
What this means for leaders
Think before you prompt. Ask for individual ideas before anyone opens the AI.
Reward the strange idea. Praise options that are unusual and possibly wrong, as well as polished ones.
Ask what is missing. After AI generates options, ask what no model would suggest.
Try this
Next time your team brainstorms with AI, add one rule. The AI's ideas are the starting point, never the shortlist.
Require at least one option the AI did not suggest. Then test it with the Experiment Planner.