When AI Stifles Critical Thinking on the Internet

In just a few years, artificial intelligence (AI) has taken over the web to the point of threatening the very diversity of ideas circulating there. This phenomenon is still largely underestimated, and its intellectual consequences could be long-lasting.

Since the advent of AI, the web has undergone a major transformation: the content it features is increasingly the product of algorithms rather than human minds. According to a joint study by Imperial College London and Stanford University published in April 2026, more than one-third of websites were generated or supported by AI in May 2025. At the end of 2022, that figure was zero.

Social media feeds are also flooded with what the English-speaking world has dubbed the slop: algorithmic content churned out in bulk to capture attention. American researcher Michelle Nie points out that the slop is specifically designed to amplify the power of digital addiction—optimized not to inform, but to keep users engaged.

According to data from the Pangram detection tools and the video editor Kapwing, nearly 40% of long-form content on LinkedIn, 30% of posts on X, and up to 60% of TikTok videos fall into this category. These figures, which come from private companies rather than academic research labs, should be interpreted with caution, but they nonetheless provide an idea of the scale of a phenomenon that remains largely obscure. 

AI, the New Educator

The success of AI on the internet stems from a structural advantage: it solves a fundamental problem of the web, namely the gap between the accessibility of information and its usability. It is easy to find content, but it can be difficult to understand it. Here, AI acts as an educator and tutor: it analyzes, defines, and structures what was previously unusable without significant effort. In doing so, it could eventually supplant the sources that have traditionally fulfilled this role: online encyclopedias, specialized forums, and media outlets.

Between 2024 and 2025, Wikipedia traffic declined by 8%, according to the Wikimedia Foundation, a significant change for one of the most-visited websites in the world. This seriously threatens the incentive to produce original content and enrich the web with new knowledge.

Standardization of Content

It would nevertheless be inaccurate to reduce this phenomenon to a flood of misinformation. The study by Imperial College London and Stanford confirms this: AI-generated content does not spread more fake news than content produced by humans. AI systems are even able to reproduce a certain stylistic variety.

The real problem lies elsewhere: this content exhibits a similarity among itself that is 33% higher than that of content produced by humans, coupled with a systematic optimism that often disconnects it from the reality perceived by readers. What AI generates isn’t fundamentally more false; it’s structurally less original and less likely to push thought to its limits.

The Austrian economist Karl Polanyi states in The Great Transformation that «the right to nonconformity would be guaranteed as the hallmark of a free society.» The hegemony of AI on the internet seems to be undermining the very conditions for this nonconformity. As content becomes more uniform, online discourse becomes less conducive to the emergence of new, evolving, and disruptive ideas—ideas that are, after all, at the heart of intellectual and political life. Such life cannot be reduced to the mere perpetuation of what already exists.

A two-tier Internet

Faced with this widespread adoption of AI, a segment of the internet is digging in its heels. Media outlets, blogs, and authors—particularly on Substack – place their content behind paywalls to shield their texts from being scraped by AI and protect their revenue. This strategy is one of the few means of resistance available to human content creators in a world where AI is moving toward a monopolistic position. The internet is thus fragmenting into two spaces: an open web, self-sustained by algorithmic output, and a paywall-protected web, where unique, embodied thought becomes a rare commodity.

The risk is that of a vicious cycle: driven by a web that they themselves have homogenized, AIs will produce increasingly shallow responses, reflecting an information space that is gradually being stripped of its substance. An increasingly intellectually impoverished digital space thus seems to be emerging, where original thought will find it ever more difficult to find a place.

A writer and philosopher by training, Baptiste Detombe is the founder of the Republican media outlet Gavroche. Every month, he takes a critical look at the technological and sociocultural changes of the time.

You have just read an open-access article published in our print edition (Le Regard Libre No. 129). Debates, analyses, cultural news: subscribe to support us and access all our content.