AI Outperforms Humans in Creativity Tests—But Only Up to a Point

Intermediate | January 27, 2026

✨ 혼자서 기사를 소리 내어 읽거나 튜터를 따라 각 단락을 반복해서 읽으세요. 레벨...

A Big Question: Is AI Actually “Creative”?

If you’ve ever used AI to brainstorm headlines, write an email, or punch up a presentation, you’ve probably wondered: Is this thing actually creative… or just fast? A large new study compared more than 100,000 people with several major AI models and found something surprising: on certain creativity tests, AI outperforms humans—at least on average. (ScienceDaily)

AI outperforms humans in creativity tests: What Was the Test?

The researchers focused on a quick psychology tool called the Divergent Association Task (DAT). It’s simple: you list 10 words that are as unrelated as possible (think: “galaxy… fork… freedom…”). Higher creativity scores come from words that are far apart in meaning, not just fancy vocabulary. (EurekAlert)

In other words: the DAT is basically a speedy “idea range” test—how quickly your mind can jump between distant concepts. That matters in real life, because the ability to connect remote ideas is a core skill behind innovation, marketing angles, and problem-solving. (Université de Montréal)

The Results: AI Beat the Average, But Not the Best

The team evaluated several large language models—including ChatGPT, Claude, Gemini, and others—and compared their scores to human results. In the DAT, GPT-4 scored higher than the human average, and GeminiPro landed close to human performance. (ScienceDaily; PMC)

But here’s the part that should calm any nervous writers: the most creative humans still win. When researchers looked at the top half of human participants, that group’s average score was higher than every AI model tested—and the top 10% pulled even farther ahead. (EurekAlert)

The “Creativity Dial”: How Instructions and Settings Change AI Output

This study also showed something very business-useful: AI creativity isn’t fixed. You can turn it up or down.

Temperature settings: higher “temperature” made AI responses less predictable and increased creativity scores in some tasks. (PMC)
Prompt strategy: telling the model to use etymology (word origins) led to higher DAT scores than using basic instructions for GPT-3.5 and GPT-4. Translation: better briefs get better ideas. (PMC)

So if AI feels “same-y,” it might not be the AI—it might be your instructions.

What This Means at Work (Without the Hype)

In a workplace setting, this supports a pretty practical takeaway: AI is a strong brainstorming assistant for average-level idea generation, but it’s not replacing top-tier creative talent yet. In fact, when it comes to quick idea range tasks, the data suggests AI outperforms humans on the average score—so teams can use it to generate options fast, then refine the best ones with human judgment. (ScienceDaily; Scientific Reports (via PMC)) In the study’s creative-writing comparisons (like haiku, story-style writing, and movie plot summaries), humans still had the edge overall—especially skilled human creators. (EurekAlert; PMC)

The smart play is collaboration: let AI generate a wide batch of options fast, then let humans pick the best angle, add taste, and make it emotionally believable.

Vocabulary

Creativity (noun) – the ability to produce original and useful ideas.
Example: The study tested whether AI could show creativity in language tasks.
Divergent thinking (noun) – thinking that produces many different ideas or solutions.
Example: Divergent thinking helps teams generate options quickly.
Participant (noun) – a person who takes part in a study.
Example: Over 100,000 participants contributed data.
Outperform (verb) – to do better than someone else.
Example: Some AI models outperformed the average human score.
Semantic (adjective) – related to meaning in language.
Example: The test measured semantic distance between words.
Model (noun) – an AI system trained to generate or analyze information.
Example: Researchers compared multiple language models.
Parameter (noun) – a setting that changes how a system behaves.
Example: Temperature is a parameter that affects AI output.
Predictable (adjective) – easy to guess; not surprising.
Example: Low temperature outputs can sound predictable.
Framework (noun) – a structured method for evaluating or organizing something.
Example: The researchers built a framework to compare humans and AI fairly.
Collaboration (noun) – working together to achieve something.
Example: Human–AI collaboration can speed up brainstorming.

Discussion Questions (About the Article)

What is the Divergent Association Task (DAT), and why is it useful?
Which AI model(s) performed strongly in the DAT comparisons?
What did the study find about highly creative humans versus AI?
How do “temperature” and prompting affect AI creativity?
What is one practical workplace takeaway from this research?

Discussion Questions (About the Topic)

What does it mean to be “creative” in your job or daily life?
When is it helpful to use AI for brainstorming, and when is it risky?
Should schools change how they teach creativity because of AI? Why?
If AI can generate ideas quickly, what human skills become more valuable?
Do you think creativity is more about talent, training, or personality?

Related Idiom

“Think outside the box” – to approach a problem in a new, non-traditional way.

Example: AI can help you think outside the box—especially when you give it a clear brief and constraints.

📢 Want more real-world English like this—short, practical, and easy to practice? 👉 Sign up for the All About English Mastery Newsletter! Click here to join us!

Want to finally Master English but don’t have the time? Mastering English for Busy Professionals is the course for you! Check it out now!

Follow our YouTube Channel @All_About_English for more great insights and tips.

This article was inspired by: ScienceDaily | EurekAlert | Scientific Reports (via PMC) | Université de Montréal