Every major communication technology in human history has had a homogenizing effect on culture. The printing press standardized spelling and grammar across languages and regional variants, while radio reduced the acoustic diversity of local musical traditions to national formats. Television standardized the imaginative narrative across half the world. And yet, each of these technologies had natural limits - geographical, linguistic, and economic - that preserved pockets of difference. Within these pockets, counter-movements could develop, and the fundamental act of individual human creation remained largely intact. As great as technological progress was, people still had to think, practice, struggle, and produce in order to move forward. What we are experiencing now, the spectacular rise of AI-driven “everything”, is structurally different. For the first time in history, a technology not only transmits or disseminates cultural products but generates them. 

And it does so on a scale, at a speed, and with a degree of integration into our everyday cognitive lives unlike any medium before it. The result is a wave of cultural standardization that is transforming our environment and, beyond that, influencing the way we think. This process is self-reinforcing and enduring to an extent that previous technological upheavals, however radical they may have seemed, never achieved.

As a growing body of research in the fields of cognitive science, cultural psychology, and AI systems analysis demonstrates, this development is already in full swing. The thesis is simple, if uncomfortable: artificial intelligence systematically narrows the spectrum of human cultural and cognitive achievements. We adapt to this restriction in ways that further accelerate it. And the window of opportunity within which this process is still recognizable and meaningfully addressed is not unlimited. What distinguishes this moment from earlier debates on technology and culture is that their consequences are not merely social or aesthetic in kind. They are biological in nature. And they strike us where we are most vulnerable as human beings: in our brains, which are naturally programmed to adapt. An environment increasingly shaped by homogenized, algorithmically generated inputs is thus quietly reshaping the very organ that was once responsible for bringing this technology into existence.

As great as technological progress was, people still had to think, practice, struggle, and produce in order to move forward. What we are experiencing now, the spectacular rise of AI-driven “everything”, is structurally different. For the first time in history, a technology not only transmits or disseminates cultural products but generates them.

A Machine that Learnt to be Ordinary

To understand what is happening to culture, it is important to know what AI systems are actually optimized for. Large language models and generative AI tools are designed to produce what users expect. They are trained on massive datasets of existing human expressions, and their outputs reflect the statistical mean of that data: what occurs most frequently, what most reliably co-occurs, and what reads most plausibly given all prior information. The result, as intended, is something that tends toward the average of human expressions rather than a specific, unique manifestation of them.

To understand what is happening to culture, it is important to know what AI systems are actually optimized for.

This tendency toward the mean is structurally determined. A study published in January 2026 by Hintze, Proschinger Åström, and Schossau demonstrated this with unusual clarity. The researchers combined a text-to-image AI with an image-to-text AI and had the two systems process each other’s results over multiple cycles, with each run starting with very different inputs. Regardless of how different the starting point was, the results approached each other. After sufficient repetition, the systems produced a narrow repertoire of atmospheric, generally appealing images. The researchers referred to this as “visual elevator music.” Over the course of the experiment, the original meaning was not only distorted but also erased through averaging.

This dynamic extends far beyond image generation. Recommendation algorithms, which now drive the majority of cultural consumption in music, video, news, and social media, operate according to a similar logic. A study published at the ACM Web Conference 2024 made an important distinction between two types of diversity: the diversity of a single user’s consumption and the differences in consumption patterns among various users. The study found that algorithmic recommendation systems primarily reduce the latter. They do not necessarily trap every user in a narrow personal bubble. Rather, they gradually and imperceptibly pull all users toward the same content. Homogenization is not individual but collective, and it occurs to such an extent that it becomes almost invisible from the inside.

Both sets of research findings paint a picture of systems that are, by their very nature, drivers of homogenization. Yet they do not even aim to eliminate cultural diversity. They merely reward what is already familiar, reproduce what is already central, and in doing so quietly push everything peripheral further to the margins. The outputs of these systems circulate. They are read, viewed, heard, and internalized. And the brain that absorbs them adapts.

Large language models and generative AI tools are designed to produce what users expect. They are trained on massive datasets of existing human expressions, and their outputs reflect the statistical mean of that data: what occurs most frequently, what most reliably co-occurs, and what reads most plausibly given all prior information.

What Happens in the Brain During this Process

The cultural consequences of this homogenization would be serious enough on their own. But this process has a second dimension that receives considerably less attention, yet is in many ways the more far-reaching one. We are not passive consumers of the environment in which we live. The human brain is a remarkably adaptable organ, and this adaptability, the very trait that enabled our species to thrive in radically different climates, languages, and social structures, is a double-edged sword. It responds to what is presented to it. And it quietly stops nurturing and storing what it no longer needs.

The neuroscientific principle at work here is well documented. Cognitive abilities that are regularly exercised are strengthened; those that are no longer used are not preserved out of mere biological courtesy, but rather they atrophy. It’s that simple. In this way, neural pathways are formed, maintained, but also abandoned. Harvard physician and researcher Dr. Aditi Nerurkar has documented how chronic cognitive outsourcing, the habitual delegation of thinking, decision-making, and problem-solving to external systems, keeps the brain in a state of functional passivity that is neurologically distinct from true rest. Our intention is to relieve the brain of a burden. Instead, we deprive it of the stimulation it needs to maintain the abilities required for conscious, independent thinking. The result is a measurable decline in the ability to concentrate, in creative range, and in the ability to sustain complex trains of thought without external assistance.

The result is a measurable decline in the ability to concentrate, in creative range, and in the ability to sustain complex trains of thought without external assistance.

Social psychologist Jonathan Haidt conducts research on the effects of algorithmically mediated environments on cognitive development. He argues that the constant availability of curated, seamless content fundamentally alters the way the brain, especially during formative developmental stages, processes ambiguity, sustains attention, and tolerates the discomfort required for deep thinking. According to Haidt, the problem is not distraction in the conventional sense. Rather, the cognitive experiences necessary for building certain skills are being bypassed: grappling with difficult material, dealing with unresolved complexity, and the patient development of a skill or an argument. What is being blocked is the very process through which competence and intellectual independence arise in the first place.

A study published at the ACM CHI Conference in 2025 made this dynamic measurable. Participants who had used AI-generated strategic frameworks for creative tasks subsequently developed less diverse ideas when working without support. Although the AI had been turned off, the narrowing of perspective had not. The study provided evidence that this was a cognitive recalibration and not merely a temporary distraction effect or a matter of convenience. The brain had adjusted its baseline expectations of what a reasonable outcome looks like and produced accordingly. A separate series of studies has shown that authors who regularly use AI writing aids produce texts that approach a statistical norm and exhibit measurably fewer stylistic and cultural idiosyncrasies than those of authors who work without such tools. The adaptation occurs gradually, below the threshold of conscious perception, which is precisely what makes it so difficult to resist.

These findings point to something more significant than a mere shift in creative habits. They suggest that the homogenization occurring in AI outputs does not stop there. It is absorbed, internalized, and reproduced by the people who consume this content. The brain is confronted with a homogenized cultural environment and shaped by it - in a way that makes it increasingly difficult to develop truly diverse and original thoughts. What begins as a means of expanding human capabilities gradually becomes, on a large scale, a force that quietly erodes those very capabilities.

In this way, neural pathways are formed, maintained, but also abandoned. Harvard physician and researcher Dr. Aditi Nerurkar has documented how chronic cognitive outsourcing, the habitual delegation of thinking, decision-making, and problem-solving to external systems, keeps the brain in a state of functional passivity that is neurologically distinct from true rest.

Two Spirals, One Direction:

What has been described so far constitutes a feedback loop. AI systems produce increasingly homogenized results, to which human perception adapts. This results in a decline in the capacity for truly diverse, independent thinking, and the cultural contributions used to train future AI systems become less diverse. Each cycle reinforces the next. This cycle requires no deliberate control. It merely assumes that the current course will continue uninterrupted.

But alongside the first spiral runs a second one, moving in the same direction at a significantly higher speed. This second spiral is about potential. AI capabilities do not develop linearly. The difference in what these systems could do five years ago and what they can do today is not a reliable indicator of their capabilities in five years. Development is exponential, and the practical implication of this argument is clear: the gap between what AI can do and what the average person can think, create, or plan strategically is widening from both sides simultaneously. While human cognitive abilities diminish through disuse and habituation, AI’s capabilities expand through continuous development. 

While human cognitive abilities diminish through disuse and habituation, AI’s capabilities expand through continuous development. 

The civilizational challenges of this process become clearer when one considers what has driven cultural and intellectual progress throughout history. The developments that have shaped human progress in science, philosophy, art, and political thought did not arise from consensus or convergence. They arose from friction: from individuals and communities who thought differently, challenged traditional ways of thinking, and created works that were truly alien at the time of their creation. This strangeness was not a side effect of progress, but the mechanism behind it. A population whose cognitive landscape has been increasingly flattened by homogenized inputs and whose capacity for independent, divergent thinking has atrophied is less capable of generating the kind of rupture that progress requires. In a study published in 2026 in “New Media and Society,” Daryani, Sourati, and Dehghani described large language models as unprecedented drivers of cultural homogenization, operating on a scale and at a speed that surpass all previous technologies. What this description implies, even if it is rarely stated directly, is that the areas most affected are precisely those that have always been the engine of civilizational renewal.

The two spirals - cognitive narrowing and technological acceleration - do not merely run parallel to one another. They also interact with one another. A population that is increasingly dependent on AI for creative, strategic, and analytical tasks provides less diverse training data for the next generation of models. These models, trained on a narrower spectrum of human inputs, produce more homogeneous results. The people who use these results adapt further. The cycle tightens ever more. And the speed at which it tightens is increasing.

AI systems produce increasingly homogenized results, to which human perception adapts. This results in a decline in the capacity for truly diverse, independent thinking, and the cultural contributions used to train future AI systems become less diverse.

The Narrowing Window of Opportunity

The moment we find ourselves in right now has a distinctive quality that deserves to be clearly identified. The process described is not yet complete. The diversity it undermines has not yet completely disappeared. The cognitive abilities it is insidiously weakening are, for the time being, still sufficiently intact so that the process can be observed, analyzed, and put into words. This essay is, among other things, proof of that. Yet it is not certain that the conditions enabling such articulation will persist. A population that has lost the cognitive habits necessary for independent, divergent thinking is also a population that is less capable of recognizing this loss, naming it, and producing the kind of analysis that could hinder its progression.

This is what distinguishes the present day from earlier phases of cultural standardization. In the past, homogenizing forces could always be countered from within the culture they were flattening, since the cognitive raw material for resistance remained available. The printing press standardized, but it also enabled the emergence of leaflets. Television flattened, but it also gave rise to the generations that turned away from it. The feedback loop currently in motion is of a different kind, for it gradually undermines the capacity for the critical distance that counter-movements require.

None of this is irreversible in an absolute sense. But reversibility is not a fixed property. It depends on time and on how much has already been lost by the time the question of reversal is seriously raised. The window of opportunity is closing rapidly. And the most honest thing one can say about the current state of this process is that we do not yet know, with any confidence, how much of it we are already too late.

A population that has lost the cognitive habits necessary for independent, divergent thinking is also a population that is less capable of recognizing this loss, naming it, and producing the kind of analysis that could hinder its progression.