MH-9569MarTech & AI
Human Pairs Beat ChatGPT Partners on Creativity, Study of 208 Finds
A randomized experiment found human pairs produced more creative, more diverse ideas than people paired with ChatGPT, and the gap widened as sessions progressed.
Wire notes
- Researchers randomly assigned 208 participants to brainstorm with either ChatGPT or another person; human pairs produced more creative ideas.
- A separate test of 12 AI chatbots found only about 42% of answer text contained the actual answer.
- Product development takes an average of 27 months from idea to launch, and up to 54 months for innovative products.
- Persona prompting increased ChatGPT's idea diversity but still fell short of human writers' range.

A randomized experiment with 208 participants found that people paired with another human produced more creative and more diverse ideas than people paired with ChatGPT on the same brainstorming task. The study, conducted by Clin Lai, a postdoctoral fellow in artificial intelligence and society at Brandeis University, and Roger Beaty, an associate professor of psychology at Penn State, adds hard evidence to a question marketers now face daily: does AI assistance help or flatten creative work?
What did the experiment measure?
The researchers used the alternate uses task, a standard test of what cognitive scientists call divergent thinking. Participants list possible uses for a common item, such as a brick or a book, and researchers score both the number of ideas and their originality. Lai and Beaty randomly assigned the 208 participants to work either with ChatGPT or with another person, taking turns.
The result cut both ways. People working with ChatGPT wrote longer responses. Human pairs generated more creative ideas, and their ideas were more distinct and diverse.
Why does AI output feel repetitive?
Individual AI suggestions could be genuinely creative. The researchers cite one example: using a book as a sound dampener. The problem was repetition. ChatGPT presented the same idea across many separate sessions, making its overall pool of suggestions homogeneous.
Other research points the same direction. One group of researchers found that repeatedly using AI to generate images produces bland, generic output, which they call "visual elevator music." Studies of AI-assisted story writing show a similar pattern: an individual writer may become more creative relative to their previous work, but AI-assisted writing across all users tends to be more similar.
The mechanics explain it. Large language models predict the next most likely words after training on massive text sets. Because their training processes resemble one another, ChatGPT, Claude and Gemini tend to share similar cadences, diction, syntax and suggestions, particularly compared with two random human writers. An AI idea may look unique in isolation. If everyone uses AI, ideas and writing style converge.
Verbosity compounds the issue. Researchers who tested 12 AI chatbots across different companies on thousands of factual questions found their answers often ran many times longer than human-written answers. The actual answer made up only about 42% of the text; the rest was padding such as repetitive information and unnecessary background.
What does this change for long creative projects?
The human advantage grew over time. The researchers measured ideas through semantic distance, which quantifies how distinct concepts are. As collaboration progressed, human pairs traversed more creative terrain, while ideas in chatbot-assisted brainstorming converged, potentially narrowing the range people explore over long stretches.
Timing matters here. Research on product development shows it takes an average of 27 months to move from initial idea to finished product, and as long as 54 months for more innovative products. Creative work in marketing is a long game, not a single prompt.
Workarounds exist but have limits. Persona prompting, in which users instruct a chatbot to respond as a specific invented character such as "a 58-year-old female founder of a construction company," increased ChatGPT's idea diversity in one study. The overall range still fell short of human writers, and other research suggests chatbots are poor simulators of human personality.
The researchers' advice is direct: chat with a person before chatting with AI. People carry idiosyncratic personal experiences that a model trained to generate average responses cannot capture. Collaborative human work also feels more meaningful, according to a separate study. For marketing teams running multi-year brand and product programs, the practical implication is to reserve AI for speed tasks and keep humans in the room where originality carries the payoff.
via digitalinformationworld.com (Source)
More from Priya Raman
Show full bio
Correspondent covering people, launches and day-to-day practice in marketing, advertising, marketing technology, media, SEO and search at Marketing Herald.
307 articles
More on the wire
- Producer J.B. Sauceda: AI Floods Comms, Great Editors Win
- 80% of Users Missed Chatbot Errors Because Fluent AI Sounds Smart
- Marketing Must Solve Its Middle Management Problem, Before AI Does
- AI Agents Won't Fix Bad Audience Data, They'll Amplify It
- 'Humanity Is Trending': Five Lessons From PRSA ICON 2026