For years, the loudest argument around generative tools has been that artificial intelligence will flatten originality and smother human imagination. A growing body of research is now pointing in a different direction, suggesting that the story is less about machines killing creativity and more about how people choose to work with them. The latest findings show that when used with care, AI can widen the range of ideas on the table, even as it raises new questions about dependence, skills and what counts as truly original work.
Instead of a simple “AI good” or “AI bad” verdict, the emerging picture is a tradeoff: systems that can accelerate and democratize creative output, while also nudging people toward safer, more average solutions if they lean on them too heavily. I want to unpack what that tension looks like in practice, and why the most surprising result from recent studies is not that AI replaces creativity, but that it reshapes who gets to participate and how.
What the new study actually measured when it tested “creativity”
Before anyone can claim that AI is helping or hurting creativity, it matters how creativity is defined and measured. In the research that has been driving this debate, creativity is usually treated as a mix of novelty and usefulness, a definition that lines up with how innovation scholars and psychologists have long evaluated new ideas. That is the same framing used in work on human and machine collaboration in business schools, where creativity is scored not just on how weird an idea is, but on whether it solves a problem in a way that others can recognize and apply.
In one recent program on human and machine collaboration, a group of scholars and practitioners gathered under the banner of Wharton Human AI Research to examine how that novelty plus usefulness formula holds up once generative tools enter the picture. Their discussion, part of an AI Horizons series, focused on whether large language models and image generators expand the space of possible ideas or subtly narrow it by steering people toward patterns in their training data. That question, how far AI stretches or compresses the creative search space, sits behind many of the most heated claims about what these systems are doing to culture.
Why some researchers say AI is expanding the creative frontier
When I look across the latest work on human and machine collaboration, one theme stands out: AI is very good at widening the first draft of the imagination. In the Wharton program on How AI Shapes Creativity, speakers described systems that can generate dozens of variations on a product concept, marketing slogan or visual style in seconds, giving teams a much broader set of starting points than they would normally have time to explore. That kind of rapid ideation does not guarantee brilliance, but it does change the odds that a team will stumble onto a surprising angle they might otherwise have missed.
The same work framed AI as both “Expanding Potential” and “Narrowing Possibilities,” a duality that captures the tension in the data. On one hand, tools that can remix patterns from across the internet give people access to a vast library of styles, metaphors and structures that used to require years of study. On the other, those patterns are learned from existing work, which means the outputs tend to cluster around what has already been done. The research suggests that the expansion effect is strongest when humans treat AI as a brainstorming partner, not an answer engine, and deliberately push beyond the first, most obvious suggestions.
The surprising boost for people who see themselves as “not creative”
One of the most striking findings in the recent literature is that AI does not affect everyone’s creativity in the same way. In controlled experiments where people were asked to generate ideas with and without machine help, Researchers found that participants who had been rated as less creative at baseline improved substantially when they were given AI generated prompts and suggestions. Their scores on standard creativity metrics rose, in some cases closing the gap with peers who had been labeled more creative at the outset.
For those already scoring high on creative tasks, the gains were smaller or nonexistent, and in some cases the machine nudged them toward more conventional answers. That asymmetry is part of why some analysts now describe AI as a kind of training wheel for imagination, especially for people who have internalized the idea that they are “not creative.” It also helps explain why advocates talk about AI as a great equalizer in the creative economy, a tool that can open doors for people who have ideas but lack the confidence or technical skills to express them at a professional level.
Evidence that AI can both sharpen and dull creative thinking
Not all of the data is flattering to AI’s impact on originality. In one widely cited experiment on writing tasks, participants who used chatbots produced work that outside judges rated as more polished and, in some cases, more creative than a control group. At the same time, the group average on measures of original thinking went down once people had access to the system, suggesting that while some individuals used the tool to stretch their ideas, others simply accepted the first plausible answer it offered. The study’s authors summed up the result bluntly: AI can boost creativity, but at a price in overall diversity of thought, a tradeoff captured in the question, Can artificial intelligence boost creativity.
For me, that result is the clearest warning sign in the current wave of research. It suggests that the risk is not that AI will suddenly make everyone’s work identical, but that it will subtly reward speed and surface level polish over deeper exploration. When a chatbot can produce a competent essay or marketing email in seconds, the temptation to accept “good enough” is strong, especially under deadline pressure. The studies hint that the most creative outcomes come when people resist that pull, using AI to generate raw material and then deliberately diverging from it, rather than treating the first output as a finished product.
How creative industries are already reorganizing around generative tools
Outside the lab, the impact of generative systems is showing up in how creative teams structure their work. A comprehensive review of professional workflows found that tools for text, image and audio generation are now embedded across advertising, design, film and publishing, with many teams using them to handle repetitive tasks like resizing assets, drafting alternate taglines or generating mood boards. According to one Comprehensive Study on how generative AI has transformed creative work, this automation is freeing up time for higher value activities like concept development, client strategy and thorough refinement of final deliverables.
That shift is already visible in job descriptions. Creative directors now routinely ask for experience with prompt engineering alongside traditional skills in storyboarding or typography, and agencies are building internal tools that sit on top of commercial models to reflect their own brand voices. The same study on Common Applications Across Creative Fields notes that beyond raw speed improvements, teams are using AI to explore more variations before committing to a direction, which can raise the creative ceiling if leaders are willing to sift through the extra options rather than defaulting to the safest choice.
Lessons from other domains where AI delivered “surprising results”
Clues about AI’s creative impact also come from fields that do not look artistic at first glance. In health research, for example, machine learning systems have been used to scan scientific literature and biological data for patterns that humans might miss. In one early project on infectious disease, scientists reported that their models had Already yielded surprising results, identifying potential treatment pathways and risk factors that had not been obvious from traditional analysis. That kind of pattern spotting is not creativity in the artistic sense, but it does resemble the process of recombining existing knowledge into novel, useful hypotheses.
For me, the lesson is that AI’s strength lies in exploring large, complex spaces of possibility faster than humans can, whether those spaces are molecules, melodies or marketing slogans. The danger is that if people treat those outputs as final answers rather than starting points, they may stop asking the next round of “what if” questions that drive real breakthroughs. In science as in the arts, the most interesting work often comes when humans take a machine generated insight and push it further, challenging its assumptions or applying it in a context the system was never trained to imagine.
Why critics warn of a “quiet extinction” of original thought
Alongside the optimistic talk of expanded access, there is a growing chorus of concern about what constant exposure to AI generated content is doing to younger minds. One essay on campus culture described a “quiet extinction of creative thought,” arguing that the abundance of AI services is changing how students approach assignments, brainstorming and even casual conversation. According to that analysis, the problem is not just plagiarism or shortcutting, but a deeper shift in habits, as people reach for autocomplete style tools before they have wrestled with a problem themselves, a trend the author links to The abundance of AI services.
The same piece notes that some of the most alarmist claims about technology and attention have been widely debunked, but it argues that generative systems are different because they do not just distract, they actively propose ideas in a person’s own voice. From my perspective, that is what makes the current moment so delicate. If students and early career workers build their thinking around machine suggestions from the start, they may never develop the tolerance for ambiguity and failure that underpins deep creativity. The research does not yet show a collapse in originality, but it does suggest that educators and parents will need to be far more intentional about when and how these tools are used in formative years.
AI as a “great equalizer” in who gets to be creative
Set against those worries is a powerful argument that AI can level the playing field in creative industries that have long been gated by geography, education and money. Advocates describe generative tools as a “great equalizer” because they give people with basic hardware and connectivity access to capabilities that once required expensive software, studio time or specialized training. One call to action framed A.I for the masses as a way to open doors, amplify voices and create opportunities for communities that have historically been shut out of media production and design.
That same theme appears in corporate research on workplace adoption, where analysts note that AI has long been introduced by its potential to be a great equalizer in productivity and problem solving. A recent study on how everyday users are closing the “ambition gap” found that people who integrate generative tools into their workflows are not just getting faster answers, they are gaining confidence to tackle more complex projects, a pattern summarized in the claim that AI has long been introduced as a democratizing force. For people who have ideas but lack formal training in illustration, music production or copywriting, that confidence boost can be the difference between staying silent and shipping work into the world.
Human creativity still has edges machines cannot fake
Even as AI systems get better at mimicking styles and structures, many practitioners argue that there are aspects of human creativity that remain stubbornly out of reach. At a recent gathering of designers and strategists, one speaker described the current landscape as a “sea of sameness,” with brands and creators leaning on similar prompts and models, which in turn produce similar outputs. The argument was not that machines are useless, but that human originality still has the edge when it comes to lived experience, emotional nuance and the ability to break rules in ways that feel meaningful rather than random, a case made vividly in a talk on Creativity That Can’t Be Coded.
I find that perspective useful as a counterweight to both hype and panic. It suggests that the most resilient creative careers will be built around qualities that are hard to quantify: taste, judgment, ethical sensitivity and the ability to connect disparate domains in ways that matter to specific communities. Machines can help with drafts, variations and even surprising combinations, but they cannot decide which risks are worth taking or which stories a culture most needs to hear. In that sense, the arrival of generative tools may actually raise the bar for human creators, pushing them to lean more heavily into what machines cannot fake rather than competing on speed or surface level polish.
Where the “AI kills creativity” narrative goes from here
Looking across the research, I see a pattern that is more nuanced than the slogans on either side of the debate. Studies from business schools, workplace surveys and controlled experiments all point to the same conclusion: AI can expand the range of ideas people consider, especially for those who start from a place of low confidence, but it can also encourage convergence on average solutions if people treat its outputs as authoritative. Analyses of The AI creativity vs Human Creativity debate argue that the future will be shaped less by what the models can do in theory and more by how individuals and institutions choose to integrate them into education, work and culture.
For now, the most surprising result is that the “AI kills creativity” story does not hold up cleanly against the data. The tools are neither pure accelerants nor automatic dampers of originality. They are amplifiers of existing habits and structures, which means the real question is whether we build norms, incentives and guardrails that reward people for using them in exploratory, critical ways. If we do, the research suggests that AI could help more people participate in creative life without flattening the peaks of human originality. If we do not, the risk is not extinction so much as a slow drift toward a culture of competent, machine smoothed sameness.
More from MorningOverview