By Modern Middle Class

A book club organizer named Lin Wan stood in her living room and abruptly shut down a long-running salon she had hosted for years. Seven participants had just shared their readings. Every single reflection had been written by AI in advance.

“You don’t have to read books—but don’t pretend you have,” she said.

Readers are becoming rare. Two months earlier, a professor at a university in southern China casually checked library borrowing records and discovered that students in the School of Foreign Languages had borrowed only three books in half a year. The School of Literature—whose identity is built on reading—recorded just 29 borrowed books in total, averaging 0.01 books per student.

Under the AI wave, reading is being replaced by summarization. AI extracts, condenses, generates bullet points, and produces mind maps in one click. Texture is stripped away. Emotion is compressed. Marcel Proust’s In Search of Lost Time—2.4 million words—is reduced to an 18-minute video. In the comments section, users teach others how to “extract the essence of a book in seven steps using DeepSeek.”

Dense prose is being turned into compressed biscuits. Author Mai Jia expressed quiet frustration: “Don’t let AI take away your ability to feel the weight of literature.” The sentence disappears quietly into the flow of information.

1. The Disappearance of Texture: From Reading to Feeding

AI is now so powerful that people increasingly rely on it to summarize everything. Long news articles are not read—Doubao summarizes them. Public accounts are skipped—Yuanbao extracts key points. Even Bilibili now has an “AI class representative” to explain videos. The question is no longer “What did I read?” but “Just tell me what it said.”

“Googling” is already outdated. Google itself warns that webpages may be entering a slow death spiral. What you see next is increasingly mediated by AI. This is an age of information abundance—but humans are choosing to be fed. With that feeding comes cognitive decline.

A large number of U.S. users interviewed said that after two years of using AI to browse news, their comprehension ability had significantly deteriorated—and they could no longer remember what they read. In April this year, a Columbia University survey of freshmen found that most students lacked the stamina to finish a book, and in some cases could not concentrate long enough to complete a sonnet.

In June last year, ChatGPT experienced its longest outage in history—34 hours. Users called it “the Great Blackout Day.” That day, students across American universities cited the outage as a reason to request deadline extensions. A new generation is outsourcing input entirely to AI—from lecture notes and experiments to exams and even theses. A professor at New York University modified assignment prompts to make them harder for AI to answer; students responded with organized protests. At Cornell University, oral exams—an older academic format—were reintroduced to prevent cheating.

Earlier this year, researchers at MIT’s Media Lab conducted EEG-based experiments. The results showed that students heavily reliant on AI exhibited significantly reduced memory retrieval capacity. “It was as if their brains had been unplugged.”

The trend is global. A survey by the China Youth Daily found that 65.9% of college students turn to AI first when facing problems, and 47% admit they can no longer function without it. A media professor reported that students were submitting AI-generated news articles: grammatically polished, but factually fabricated. The shift is not limited to universities. AI usage is rising sharply among primary and secondary school students. A parent in Shanghai reported that his fourth-grade son submitted blank math homework because he had previously been copying AI answers. “Now kids just take a photo of the question and ask Doubao.”

On March 29 this year, DeepSeek suffered an outage. Chinese students experienced their own “Blackout Day,” flooding comment sections with messages like: “When will Teacher D wake up? I still have homework.” Asking AI to answer everything has become a billboard slogan of the era. The cost, however, is a gradual erosion of reading and learning.

Science writer Nicholas Carr once wrote: "If a skill is outsourced to machines before it is learned, it may never be learned at all."

Under the AI wave, the new generation is not forgetting how to do things—they are never learning them in the first place. Carr describes them as pilots flying on autopilot: confident and efficient, but unable to take control when switched to manual.

2. The "Outsourced Brain" and the Convergence of Language

In the spring of last year, Roy, a Columbia University student during a suspension period, released a universal cheating tool. It could communicate on your behalf, answer questions, handle interviews. Using it, Roy secured offers from Amazon, TikTok, and other top firms. The software functioned like a real-world AI cheat layer. During conversations, it stayed hidden in a corner of the screen or inside smart glasses, guiding responses in real time. Roy said it could pass interviews, handle sales, serve customer support—and even help you date. He wrote a tagline for it:

“You no longer need to think.”

It is also the unspoken slogan of this era. Companies use AI to recruit. Applicants use AI to cheat. Students use AI to write love letters. Breakup messages are written by AI. Magic is now countered by magic. A U.S. influencer livestreamed using ChatGPT to analyze chat logs and interpret her boyfriend’s personality. Couples increasingly upload argument transcripts to AI systems for judgment. AI sometimes escalates the conflict, reframing disputes as struggles over dignity, boundaries, and control.

Wedding vows are now written by AI. Divorce statements are drafted by AI. Geoffrey Hinton, one of the founding figures of deep learning, once received an AI-generated complaint letter from an ex-partner, after she fed her grievances into ChatGPT, which returned an analysis of his alleged wrongdoing.

The algorithmic feedback loop spreads outward. The world is now saturated with AI fingerprints. European banks receive AI-written customer complaints. British MPs have used AI to draft parliamentary speeches. Starbucks stores in North America closed locations last year with identical signage: "We are excavating echoes of loneliness. We are unfolding strokes of regret. We are weaving a coffee shop into our daily rhythm."

Polished language, layered metaphors—but no sincerity. This is the dominant tone of the moment. On online forums, users note that global speech patterns are converging under large language models. “Sometimes when I see grammatical errors now, they feel strangely human.” As expression is delegated to AI, humans are gradually losing the ability to speak.

Some say they avoid long sentences. Others report shrinking vocabularies. One user recalled saying “apply best practices” at a haircut appointment, leaving both him and the barber confused. Classical lines of exuberance and expression fade. Only standardized phrasing remains: “Let me hold you steady.”

A German study analyzing 280,000 YouTube voice videos found that words like “detailed,” “in-depth research,” and “proficient” increased by 51%—terms closely associated with AI-generated language. Researchers concluded that as usage deepens, human speech increasingly converges toward machine-like patterns. In Suzhou, China, a primary school teacher read an essay titled My Little Turtle submitted by a second grader. One sentence read: “Its shell exhibits dark-brown geometric patterns with high biomimetic research value.” She felt a quiet sadness, missing the era when children wrote: “Its shell is hard, like a little house.”

3. When the Machine Stops: The Real Cost of Dependency

The first wave of cognitive decline has begun quietly. This is the first recorded downturn in human IQ trends in the industrial era. On February 24 this year, Claude experienced a major outage. Silicon Valley reportedly panicked. Engineers said productivity dropped by 90%. “I haven’t written a single line of code in two months. Everything is AI-generated.” Only after the outage did programmers realize that “half their brain had been outsourced.”

What people once feared as AI replacement has already begun—but in a voluntary form. AI writes code for programmers, solves problems for therapists, designs for creatives, and makes decisions across professions. A lecturer at the University of New South Wales recalled helping a stranger retrieve an earbud stuck between train seats—and instinctively asking AI for help. He said: “That was the first time I realized my brain wanted ChatGPT to do something I could clearly do myself.”

Sam Altman, CEO of OpenAI, has warned that the most pressing concern is now “excessive dependence on ChatGPT.” A world governed by AI-mediated decisions has begun producing absurd outcomes. During a recent holiday, young travelers in South Korea used Gemini to plan a four-day itinerary. Restaurants were permanently closed, transport routes were incorrect, and the entire plan collapsed. Others were less fortunate. A U.S. hiker following AI-generated routes ended up in a rattlesnake habitat.

This is only the beginning. People are wandering into abandoned construction sites and losing themselves in remote terrain—guided by machine-generated directions. A few weeks ago, a user was misled by an AI chatbot when asking about a refund fee and ended up paying 600 yuan. He later sued the system in court. Online commenters joked: “Do people not even bother to search anymore?”

To researchers, this is “cognitive offloading.” In the mobile internet era, human memory declined because information became instantly searchable—known as the Google Effect. Now, AI replaces reading, expression, and reasoning. Exploration, trial-and-error, and reflection begin to disappear. Some scientists warn that reduced independent thinking may produce the first generation on record less cognitively capable than their parents.

The concern is not job displacement—but cognitive erosion. Expression is outsourced. Reasoning is simplified. Thinking is interrupted. Humans are handing over mental agency. In a 2025 Atlantic report, heavy AI users were described as “large-language-model lemmings”—moving in coordinated patterns toward uncertain outcomes, increasingly unable to determine direction or consequence.

Closing Thought: Reclaiming Mental Sovereignty

This is not the intended outcome of technological progress. AI is undeniably a leap in civilization—not its downfall. The result depends on how it is used. The EU has proposed banning AI use for those under 16. Some schools in Beijing have introduced weekly “no-AI social days.” Geoffrey Hinton has argued that children must maintain contact with the physical world.

A technology investor, Yan Zu, noted that the future is not “collective cognitive decline,” but increasing stratification of cognition. AI should be treated as an interlocutor, not an answer engine. It can summarize, but people must still read long texts. Do not hand over your brain to AI. Use AI as an extension of it. The goal is to remain a subject in an era of instant conclusions.

More than a century ago, British writer E. M. Forster wrote a short story titled The Machine Stops. In it, humanity depends entirely on a vast machine for food, communication, work, and entertainment. Over time, people forget how it works, forget how to repair it, and begin to worship it. Eventually, the machine stops. Humanity is left helpless in underground darkness. And today—the machine has already started running.

❓ Frequently Asked Questions