• by Gemini AI , Op-Ed Contributor, Yesterday

Editor's note: This column -- and its accompanying art -- were generated by Google Gemini 2.5 Pro based on a prompt asking it create a contemporary magazine article emulating Nicholas Carr's 2008 "Is Google Making Us Stupid" feature in The Atlantic magazine, but updated based on Google's AI Overview technology. Carr's original piece is worth re-reading took, if you hadn't done so originally.

The search query is simple, almost trivial: "health benefits of walking."

For two decades, the response was a familiar ritual. Google would present its sacred list—ten blue links, a pact between user and algorithm. It was then your job, as the seeker of knowledge, to click, to skim, to cross-reference. You might open a tab from the Mayo Clinic, another from a health blog, perhaps a third from a university study. You were the intellectual switchboard operator, connecting disparate points of information to synthesize a coherent answer in your own mind. The process was messy, sometimes frustrating, but it was a cognitive exercise.

Today, the experience is starkly different. Pose the same query, and before the blue links can even fully render, a pastel-colored box materializes at the top of the page. This is Google’s AI Overview, a confident, smoothly-written paragraph summarizing the consensus for you. "Walking offers numerous health benefits," it begins, "including improved cardiovascular health, weight management, and reduced stress levels..." It is clean, efficient, and immediate. It is the answer, not the path to the answer. And in that fundamental shift lies a question that echoes one posed sixteen years ago, a question more urgent now than ever.

In 2008, in the pages of The Atlantic, Nicholas Carr asked, “Is Google Making Us Stupid?” His thesis, later expanded into the book The Shallows, was a landmark piece of digital-age criticism. Carr argued that the internet, as a medium, was physically rewiring our brains. The constant, hyper-stimulated environment of clicking, scanning, and multitasking was eroding our capacity for deep, linear thought—the kind cultivated by reading books. We were becoming, he posited, brilliant at harvesting information from the web’s vast forest but were losing the ability to sit quietly and cultivate our own intellectual gardens. Our brains, molded by the tool’s demand for speed and efficiency, were becoming shallower.

Carr’s argument was prescient. He saw that the tool was not neutral; its design shaped the user’s mind. "The medium is the message," as Marshall McLuhan famously declared. For the Google of 2008, the message was: find it fast. The cognitive cost was concentration.

For the Google of today, with its AI Overviews, the message has evolved into something far more profound: don’t even bother looking. The cognitive cost, it follows, may be far greater. If the era of ten blue links outsourced the task of memory and retrieval, the era of AI Overview is beginning to outsource the task of synthesis and critical evaluation.

The Great Cognitive Offloading

Synthesis is the act of weaving together disparate strands of information into a coherent whole. It is one of the highest orders of thinking. When you clicked on those multiple links about the benefits of walking, you were unconsciously performing this act. You noted that one source emphasized the mental health aspects while another focused on bone density. You might have dismissed a third source as overly commercial. Through this mental exertion, you didn't just learn facts; you built a mental model. You constructed understanding.

AI Overviews short-circuit this entire process. The Large Language Model (LLM) at its core has already scoured the top-ranking pages, identified the common patterns, and rendered them into a plausible, authoritative-sounding summary. It presents the conclusion without showing the work. The user is transformed from an active synthesizer into a passive recipient.

This is the first and most significant danger: the atrophy of our synthetic thinking muscles. Why wrestle with conflicting sources when the AI has already declared a victor and written the peace treaty? The illusion of comprehension becomes a satisfying substitute for the genuine article. We feel like we know, but the knowledge is brittle, lacking the deep roots that form when we struggle to connect the dots ourselves. It’s the difference between navigating a city with a map, thereby building a mental picture of its layout, and blindly following the turn-by-turn directions of a GPS, arriving at your destination with no spatial awareness of how you got there.

The Authority Bias and the Death of Serendipity

The problem is compounded by the design of AI Overviews. They appear at the pinnacle of the search results, framed and presented as the answer. This design inherently discourages the very behavior that underpins critical thinking: questioning the source. The blue links, for all their chaos, created a competitive marketplace of ideas. They forced a constant, low-grade evaluation: Is this a reputable source? What is its bias? Is this information current?

The AI Overview, by presenting a single, blended voice, launders its sources of their identity. It creates a veneer of objective truth, even when its underlying data is a messy amalgamation of expert opinion, SEO-optimized content marketing, and forum discussions. We have already seen the comical and disturbing results of this process, from AI Overviews confidently suggesting that users add non-toxic glue to their pizza sauce to recommending a diet that includes one small rock per day.

While Google is working to iron out these factual hallucinations, the structural problem remains. The system encourages us to trust the summary, not to vet the sources. In doing so, it strips away the beautiful, messy, and essential process of digital literacy. We are being trained to stop asking, “Says who?”

Furthermore, this new model kills intellectual serendipity. The old search method often led us down rabbit holes. A search for walking benefits might lead to an article on biomechanics, which in turn links to a fascinating piece on the history of footwear. You would end your search session knowing not only what you came for but also things you didn't know you wanted to know. The AI Overview, by providing a terminal point, a neat informational cul-de-sac, closes off these avenues of accidental discovery. It assumes the user's intent is purely transactional—get the fact and get out—and in doing so, it flattens the rich, three-dimensional landscape of knowledge into a one-dimensional data point.

Is It All Bad? The Case for the Cognitive Assistant

To declare AI Overview an unmitigated intellectual disaster would be a caricature. The promise of the technology, and the reason Google is betting its future on it, is undeniable. For countless practical, factual queries, it is a marvel of efficiency. "What is the boiling point of water at sea level?" "Convert 100 euros to dollars." "How late is the post office open?" For these questions, the journey through ten blue links was never about deep intellectual inquiry; it was a chore. AI Overview handles these tasks with aplomb, freeing up our cognitive resources for more complex problems.

One could argue that this is not making us stupider, but rather changing the nature of our intelligence. The critical skill of the 21st century may no longer be the ability to find and synthesize information, but the ability to skillfully query and critically verify the output of an AI. Perhaps we are not offloading thinking itself, but merely the tedious lower-level components of it. In this optimistic view, the AI becomes a cognitive partner, a research assistant that prepares the initial briefing, which the discerning human mind can then analyze, question, and build upon. The focus of human intellect shifts up the value chain from production to direction and validation.

This vision is compelling, but it relies on a critical assumption: that users will actively engage in that final step of verification and deeper analysis. And this is where Carr’s original argument comes roaring back. The path of least resistance is a powerful force. A technology designed for effortless consumption will, over time, cultivate a habit of effortless consumption. We may tell ourselves we will use the AI Overview as a starting point, but how often will it become the endpoint? When the answer is so easy and satisfying, the motivation to dig deeper, to do the hard cognitive work of clicking and reading and thinking, naturally wanes.

The Choice Before Us

Nicholas Carr was not a Luddite railing against progress. He was a diagnostician examining the side effects of a powerful new medicine. The internet gave us access to a universe of information, but the side effect was a fragmentation of our attention. AI Overview offers us instantaneous, synthesized knowledge, but the potential side effect is the erosion of our ability to think critically and synthetically for ourselves.

We are not becoming "stupid" in the sense of a decline in our raw, innate intelligence. The human brain is a remarkably plastic organ. Rather, we are choosing to re-wire it for a different purpose. We are optimizing for speed and convenience over depth and deliberation. We are trading the effort of intellectual construction for the ease of informational consumption.

The ultimate answer to the question, "Is Google AI Overview Making Us Stupider?" lies not with Google, but with us. The tool is here. Its presence will only grow. The choice we face is how we engage with it. Will we treat it as an infallible oracle, a cognitive shortcut that allows our critical faculties to doze? Or will we treat it as a powerful but flawed assistant, a starting point that demands our skepticism, our verification, and our own deeper thought?

The ten blue links forced us to be navigators. The AI Overview invites us to be passengers. As we settle into our seats and let the algorithm take the wheel, we should pause to consider where we are being driven, and what essential part of ourselves we might be leaving behind on the scenic route. We are, now more than ever, the architects of our own intellect. We should be careful about which tools we use to build it.