What LLMs Are Quietly Doing to Creativity
How AI makes it harder to think differently.
Posted May 8, 2026 | Reviewed by Tyler Woods
What helps us be creative? For years, researchers pointed to divergent thinking , or the ability to think differently from one another, to follow unexpected threads, to sit with not-knowing long enough to surprise yourself. Now a pair of 2026 studies suggest emerging research suggests this might be happening. And the culprit isn't distraction or busyness, but rather the tools we've started thinking with.
I'm a speech-language pathologist and the author of The Art of Talking with Children. My work centers on what I call Rich Talk , or conversations with children that are adaptive, open-ended, and deeply curious. And what I'm seeing now, both in research and in my clinical work, concerns me. That talk, and the creativity that underlies it, feels deeply at risk.
The first study, published in PNAS Nexus in March 2026 , examined creativity outputs across multiple large language models and compared them to a human's response on standardized creativity tasks. The authors found that LLM responses mirrored other LLM responses far more than humans mirrored other humans. As the authors concluded, using LLMs as creative partners "may drive users toward similar 'creative' outputs," no matter which specific LLM we're using. Even if we change from one LLM to another--for example, jumping from Claude to ChatGPT--we're likely to end up with a slightly different version of the same response.
The second study, published in Trends in Cognitive Sciences , moved from outputs to the mind itself. The researchers argued that the issue is far more profound--and that the LLMs may be standardizing even how we think. "Individuals differ in how they write, reason, and view the world," one author Zhivar Sourati writes. "When these differences are mediated by the same LLMs, their distinct linguistic style, perspective, and reasoning strategies become homogenized."
The homogenization problem in LLMs is, fundamentally, a conversation problem. It's about the back-and-forth that we depend on for creative and generative thinking getting lost.
Recently, I've been noticing a pattern in AI -generated content. In reading AI-generated ads and marketing language, I kept seeing words like "genuine," "authentic," "truly," "meaningful." Finally, I asked the AI directly: why do you keep using the word "genuine"?
The response was telling: "When I suggest 'genuine dialogue,' I'm using language that signals consciousness, emotion , human connection--all things I don't have. When a system can't actually care, it overproduces the linguistic markers of those things."
This is the issue, ironically identified by an AI: output without emotion, and words without the underlying care. If we don't start to take a more critical view, that's what we're teaching children to produce--and, ultimately, to consume.
Rich Talk is characterized by the ABCs: it's Adaptive (shifting with the person or people around you), Back-and-forth (not a one-way lecture), and Child-driven (focused on the needs and wants of others as well as your own). This kind of talk fundamentally requires empathy, including perspective-taking . When we have output without emotion, this talk cannot work well. We have words, yes, but we lack the underlying care.
The very capacities that Rich Talk builds — curiosity, comfort with uncertainty, the ability to follow an unexpected thread — are the ones that homogenizing tools erode. Rich Talk isn't about transmitting information efficiently. It's about the messy, unpredictable back-and-forth that helps children learn to think for themselves. It's about asking questions you don't already know the answer to. It's about following a child's strange idea past the point where it would be easier to redirect.
When we outsource our thinking to a tool that predicts the most likely next word, we're not just getting help. We're subtly training ourselves-- and potentially our children-- to value the average idea over the original one. They smooth out their language, and with it, their inner thoughts.
This matters especially for children, who are still developing their inner voices. A child who learns to reach for an AI assistant every time they're stuck might struggle to develop the capacity to sit in the discomfort of not-knowing, and then find their own way through. That discomfort is precisely where original thinking begins.
A child I worked with showed me how they wrote a book report using Alexa. "I just have to ask them what the book was mostly about," they said, "and then I listen to that a couple of times and write the main points down. When I am supposed to write about how I felt about the book, I just ask if they feel good or bad about it--and what I should say." Alexa works differently under the hood, but the dynamic is the same: the child was learning that answers come from outside, not from the messy process of figuring out what they themselves think. They weren't only failing to develop their own opinion. They were absorbing a deeper lesson: that creativity comes from the machine, not from the iterative process of thinking or talking it out.
Of course, my own evidence is anecdotal. It echoes what I have heard from colleagues, families, and researchers who speak about patterns with their own kids. But we need far more studies over months and years to understand the true nature of these changes, and whether creativity might be developing, but simply in different ways.
For Rich Talk, we need to trust our own individual minds. We also need to rely, at times, on our intuition : when to stay silent, when to speak out, what another person's body language means. All of this allows for creativity, precisely because we are human and responsive. The divergence in human thought is, fundamentally, a good thing. It allows for genuine dialogue, not only an echo chamber. And if that's what we are losing, we may be losing the foundation of human creativity.
This isn't an argument against technology. It's an argument for protecting the kinds of conversations that help children become themselves, while also teaching them to engage with technology on their own terms.
Start by leaving space. When you talk with a child, resist the pull to fill silence, redirect strange ideas, or smooth over uncertainty. The moment a child says something that intrigues, surprises, or confuses you may be the moment worth following. Ask: "What do you mean by that?" The question is about following the thread of a thought, understanding the child's felt sense.
Talk about the tools honestly. Even young children can understand that an AI doesn't think; it pattern-matches. It guesses what word probably comes next. That simple explanation is a gift. It lets children use technology without mistaking it for a mind and without outsourcing the question of what they themselves think.
And finally, help kids experiment with LLMs critically, rather than simply avoiding them. Ask them, for example, to request the same story told for a third-grader, an adult, or a baby. Then talk about what the LLM added and what it left out. Why did it make those choices? What assumptions is it making about what a baby or a third-grader needs? This kind of thinking helps kids be creative around technology: noticing the patterns, questioning the outputs, staying in charge of their own minds.
The research now confirms why protecting divergence matters. LLMs are pushing toward the average. Our conversations with children can pull in the opposite direction: toward the specific, the surprising, the unrepeatable. In a world that's quietly standardizing how we think and speak, the most radical thing we can do with our children is keep talking differently.
Share this post Facebook Bluesky Linkedin Email
There was a problem adding your email address. Please try again.
By submitting your information you agree to the Psychology Today Terms & Conditions and Privacy Policy
Rebecca Rolland, Ed.D. , is a speech pathologist and lecturer at the Harvard Graduate School of Education and serves on the faculty at Harvard Medical School. She is the author of The Art of Talking with Children (HarperOne).
Get the help you need from a therapist near you–a FREE service from Psychology Today.
This article is part of the Bringwise Psychology Journal — daily insights on human behavior, mental health, and personal growth.