How creative is AI? Right now

I have been talking about creativity intensively for close to a year, from gearing up for the launch of The Creativity Choice to following up on the interest it generated. Whether a podcast or a speaking gig, almost invariably the second to last question ends up being about AI. (The last one tends to be something of a summing up and imparting parting words.) At this point I find the regularity funny. But I also think this is a very important question worth answering step by step.

So, here is my take on creativity and AI. With an important note that I am talking at the beginning of 2026 and I am talking about LLM-based AI tools. I am not talking in 2336, for instance, when according to Star Trek history android Data is created. I am not going to pretend that I have the ability to foretell the far away future. I am curious and excited about there one day being a Data, but also apprehensive (if you know Star Trek, you know Data has a ‘brother’ Lore).

Setting aside fears of robot overlords, here is where we are right now. For a conversation that directly prompted me to write this out, here is a video from a recent chat with Mark Blackwell on the Arkaro podcast.

As soon as generative AI tools came on the market, researchers started conducting studies on creativity of ideas they produced. These studies gave the same kinds of tests of creative idea generation to AI and groups of humans. Some prominent examples. In the Alternate Uses Task, the goal is to come up with ideas for different potential uses for everyday objects (e.g., brick); the Consequences Task asks for ideas about what would happen if unlikely or fanciful hypothetical scenarios transpired (e.g., if humans no longer needed sleep); the Divergent Associations Task prompts respondents to come up with 10 nouns as different from each other as possible (excluding proper nouns). Answers are then evaluated for originality.

Studies have found that on average, AI tools generate more original responses than human participants. This is a reliable finding.

But I think it is important to consider these results a bit more closely.

It is true that scientists often compare average performance of different groups, such as AI and humans in this case (or artists vs. scientists or men vs. women etc.). However, we can also ask different questions.

When we ask whether AI tools are more creative than creative people, we get a very different answer. Yes becomes a no. If data are analyzed to compare not averages, but most creative responses of AI and humans, the answer is that AI is not coming on top.

I have heard some critics say this kind of analysis is massaging of results when you don’t like what common analyses show. And I have to disagree. This is defining creativity in a way that we define it in everyday life. We are not impressed by averages of ideas that teams come up with in their brainstorming sessions. Rather, we are impressed by the most creative ideas that come from those sessions (and end up being developed further and transformed into products or performances). The highest creativity that is possible is what makes a difference in the world.

Contemporary scientific research often deals with averages and examines performance of people from the general population, rather than recruiting and studying those who show evidence of creativity in what they do in the real world. The reasons for this are many and many are practical in nature, driven by pressures of academia (if you are going to take time to recruit designers or engineers or scientists, you will not be able to publish as many papers!). Perhaps this a topic for its own post if it might be of interest. Research of creativity in not particularly creative people is one of my biggest critiques of my field.

But much of foundational historical research on creativity was not built on samples of college students or people who complete surveys for extra money. Frank Barron, one of the founders of creativity studies, defined creative individuals as those who were in the top 15% for all eight different measures of creative thinking and in the top 2% for at least two of them. Donald MacKinnon and the team at the University of California at Berkely studied people who were nominated as creative by their colleagues (e.g., architects, mathematicians etc.). And Mihalyi Csikszentmihalyi in his monumental work studied Nobel prize winners and other similarly highly creative individuals.

Even if AI does not produce the most creative ideas when compared to people who have developed their creative potential, it does not mean that it is not useful.

This is the case when writers, for instance, consult an AI about character development or plot elements. Even if they do not end up using the exact words produced by the machine, the exercise might be helpful. It can provide leads that become inspiration. Or a human creator can take an idea and then transform or express it in their own style and by adding the lived experience.

A friend in a recent LinkedIn post expressed it best. Consulting AI is helpful to check for obvious ideas. AI is great at the obvious (and humans sometimes/often are not aware of the obvious). It is great at it because it is based on what already exists in a corpus of work floating out in the culture. It can retrieve it and by consulting it we can make sure we are not missing something that we should not miss. That is truly helpful in creative work.

Full disclosure, it did not occur to me to consider this question on my own. It was a question I was asked in an interview for a business magazine in Korea. And it got me going.

I don’t think that AI will fundamentally change the nature of human creativity. It will become a tool that supports it, whether as a partner in searching for inspiration, doing background research, or assisting with formal aspects of creative work.

In my opinion, two big things are likely to change. One is a societal difference in relevance given to creativity by many organizations. When humans were the only ones who could generate something new, even when the explicit goal was not necessarily creativity, some creativity could happen by sneaking in. Now, businesses that strive for efficiency above all will likely conclude that AI ideas are good enough. If close to full automation becomes their choice, bursts of highly creative products might become more rare. Something highly creative will be pushed into the domain of those who are willing to not only be good enough, but be set apart.

Another way I expect that evolution of AI will affect human creativity is more optimistic and psychologically more interesting. I think it will bring to people’s awareness the importance of identifying and framing problems. Decades of research show that people come up with most creative solutions when they spend much of their time and effort on examining the problem from different perspectives, whether they are creating a still life painting or solving work-related problems. AI tools can generate ideas, but they are not sensitive to problems or have agency to ask questions. As we see that we are not the only ones who can generate ideas, we will realize to a greater extent what we have to do and do well. In interacting with the world, we notice what works well and what does not, and use those observations and feelings as inspiration for what problems to pursue. Frustrated by the grocery shopping experience? It is an inspiration to ask how it could be different and a motivation to solve this problem (by starting a company like Instacart).