Is Gen Z really the dumbest generation?

Before you ask an AI whether this title is even talking about you, here is the short answer. Generational boundaries are not laws of nature, and they vary slightly depending on the source. For this article, I use the now common classification in which Gen Z was born roughly between 1997 and 2012.

If you were born between 1997 and 2012, this really is about your generation. If not, you still cannot sit back and relax. The problem discussed here does not stop at a birth year.

The provocative claim that Gen Z is the dumbest generation in modern history strikes me as containing a kernel of truth. Still, “dumb” is not a clean scientific measurement. It is a provocative headline for something that can be described in a more nuanced and equally troubling way: important academic and cognitive performance is declining while digital systems offer us ever more opportunities to avoid strenuous thinking.

Where the claim comes from

In January 2026, neuroscientist and former teacher Jared Cooney Horvath told a US Senate committee that Gen Z was the first generation in modern history to underperform its predecessors on virtually every available cognitive measure. He cited attention, memory, reading, mathematics, executive functions, and even general IQ.

That is a strong statement. Social networks and media quickly reduced it to the claim that Gen Z had now officially been confirmed as the dumbest generation. It is not that simple. Horvath offered an expert assessment at a political hearing. There is no single study or scientific authority that has officially given an entire generation this label.

Nor has the cause been conclusively proven. Horvath sees a major break around 2010 and primarily blames the widespread introduction of digital devices in classrooms. Other explanations also belong in the discussion: pandemic school closures, social inequality, changing curricula, grade inflation, test-optional admissions, psychological stress, and, of course, the way smartphones compete for our attention.

The provocative wording is debatable. The warning signs beneath it are much harder to dismiss.

The numbers are getting uncomfortable

The US National Assessment of Educational Progress, NAEP, shows a clear decline among twelfth-grade students. In 2024, 45 percent performed below “NAEP Basic” in mathematics, compared with 40 percent in 2019. In reading, 32 percent were below “Basic,” compared with 30 percent in 2019 and 20 percent in 1992. These categories are not the same as normal grade levels, and NAEP explicitly says they should be used with caution. The trend remains bad, especially among the weakest students.

The problem becomes even more tangible at the University of California San Diego. A university report says that in 2020, about 30 incoming students had mathematics skills below high-school level. Five years later, there were more than 900, nearly thirty times as many. About 70 percent of that group were even below middle-school level. Many struggled with fractions and basic algebra.

The original report was at times summarized as saying that one in eight freshmen was below middle-school level. That wording was later corrected. The real figures are bad enough. There is no need to make them worse.

Another detail is particularly revealing: more than a quarter of the students placed into the university’s lowest mathematics course in 2024 had previously earned a 4.0 GPA in high-school mathematics. Top grades and actual preparation no longer matched. Anyone who receives good grades for years despite missing the fundamentals will understandably consider themselves more competent than an independent assessment shows them to be.

Reading does not look any better. A YouGov survey of 2,203 US adults found that 40 percent had not finished a single book in 2025. The median was two books. People aged 18 to 29 averaged 5.8 books. Audiobooks were included. This does not prove that young people never read, since the internet and messaging also consist of text. A book, however, requires something a feed rarely does: staying with one train of thought for an extended period.

The feed never asks follow-up questions

The endless feed existed long before ChatGPT. TikTok, Instagram, YouTube Shorts, and similar formats were not built so that we would feel satisfied and stop after twenty minutes. They are designed to trigger the next movement of the thumb.

That does not automatically change IQ. But it trains a particular way of dealing with information. The algorithm selects the next topic, supplies the stimulus, and replaces it seconds later. I do not have to formulate a question, choose a source, develop an objection, or even decide what I want to know next. I merely receive.

A 2025 survey commissioned by Sprout Social among more than 2,200 social-media users in the United States, the United Kingdom, and Australia found that 41 percent of Gen Z first turn to social platforms when searching for information. Traditional search engines came in at 32 percent and AI chatbots at 11 percent. This is a marketing study, not neutral educational research. But it fits behavior that is easy to observe in everyday life: research is increasingly confused with consuming a short video.

A short video can be excellent. It can spark interest, make a complicated issue visible, and offer an entry point into a subject. It becomes a problem when it is both the beginning and the end of the research. Anyone who never opens the primary source will not notice when “45 percent below NAEP Basic in mathematics” suddenly becomes “almost half cannot read.”

Does this consumption actually damage the brain?

The term “brain damage” is tempting but scientifically too crude. There is currently no solid evidence that social media or short-form video permanently causes physical damage to a healthy person’s brain or automatically lowers IQ. There is, however, growing evidence that excessive and highly fragmented consumption can impair attention, memory, self-control, and the processing of longer contexts.

An experiment at Ludwig Maximilian University of Munich studied 60 people who used TikTok, Twitter, YouTube, or no media during a memory task. Only the TikTok group showed a clear deterioration in the ability to resume and carry out an intention formed before the interruption. The researchers did not simply attribute this to “social media,” but to the combination of short videos and rapid context switching.

A 2026 study published in npj Science of Learning went one step further. Fifty-seven young adults watched either one coherent ten-minute video or seven short, comparable segments. The short-video group remembered significantly less afterward. fMRI measurements also showed lower activation and weaker functional connections in networks involved in information integration, cognitive control, and memory retrieval.

That is a measured difference in brain activity during a specific task. It is still not proof of permanent damage. This study was also small, examined only young adults, and cannot show what happens over months or years. A meta-analysis of problematic screen use found small to moderate deficits across 34 studies, particularly in attention and executive functions. Every included study, however, was cross-sectional. The crucial question therefore remains open: does excessive media consumption cause the problems, or do people who already have attention problems use these media more often? Both directions probably reinforce each other.

I therefore do not want to condemn social media as a whole. A group chat, a specialist forum, a long explanatory video, and an hour in an endless feed are not the same activity. What matters is whether I am actively communicating and processing something or whether an algorithm takes over my attention every few seconds.

In a short video, Elon Musk is asked which innovation has made us worse rather than better. His answer is short-form video. Musk’s statement is a personal opinion, not a study. Yet it fits remarkably well with the pattern that research is slowly making visible. It is especially ironic that this warning reaches us as a short-form video.

Mostly American data, not a global ranking

At this point, the article needs a geographical boundary. The figures on NAEP, UC San Diego, book consumption, college graduates, and wealth primarily describe the United States. The online entrance-exam case comes from Mexico. None of this supports a global ranking of generations or the claim that every young person in the world is developing in the same way.

China is not a country without social media. Many Western platforms are blocked, but domestic services such as Douyin and Kuaishou are enormous. According to the China Internet Network Information Center, China had around 1.04 billion short-video users in 2024, with average daily use of 156 minutes. At the same time, China has introduced much stronger youth modes, time limits, and content controls for minors. This does not make the country a media-free control group. It shows that the same technology is used under different platforms and rules.

Interestingly, research from China is not generally absent. A meta-analysis published in 2026 on short videos and mental health found that 55 of its 58 included studies were conducted in China. That creates the opposite problem: findings from Chinese samples cannot automatically be applied to the United States, Europe, or Africa either.

Africa really is heavily underrepresented in this research. A review of social media and depression among adolescents found no African sample among 34 publications. Europe was represented, but with its different school systems, languages, social protections, and smartphone rules, it is not a uniform comparison region either. The honest conclusion is not that America has been proven dumber or China automatically smarter. We see most clearly where measurements are taken, and we are far from measuring the same things everywhere.

Intended as satire, now almost a documentary

All of this repeatedly makes me think of Idiocracy. Mike Judge’s film was released in 2006. In it, a completely average American wakes up after 500 years in a society where consumption, advertising, entertainment, and convenience have displaced almost every form of competence and reasonable decision-making.

I have probably watched the film three times over the past 20 years. Each time it felt slightly less like an exaggerated comedy and slightly more like an uncomfortably accurate prediction. Of course, Idiocracy is not a documentary. The biological explanation with which the film introduces its dumbed-down future is highly simplistic and is not my thesis either. Something else is frighteningly accurate: a society does not have to lose its intelligence all at once. It is enough to reward attention, education, and expertise less and less while the easiest stimulus is always available.

The film gets a clear recommendation from me. Idiocracy arrived in cinemas as satire in 2006. If we keep moving in the same direction, perhaps it can soon be called a documentary. That is the disturbing joke: the film may simply have appeared twenty years too early.

Then AI arrived

Social media fragmented attention into small pieces. Generative AI now also offers to take over the strenuous parts of thinking entirely.

I notice this in myself. As soon as a text gets long, two documents need comparing, or a sentence does not work immediately, reaching for AI is tempting. It can provide an initial overview, identify differences, or help with wording. It only becomes a problem when the summary replaces the source and the suggested wording replaces one’s own thought.

That convenience is precisely what makes the tool so useful. And it is precisely what makes it dangerous.

An MIT research group asked 54 people to write essays under three conditions: using only their own minds, using a search engine, or using a language model. EEG measurements showed the strongest and most widely distributed functional connectivity in the unaided group. The search group fell in the middle, while the LLM group showed the weakest connectivity. Participants using an LLM were also less able to quote from their own text and felt less personal ownership of it.

That sounds like definitive proof that ChatGPT makes the brain wither. It is not. The work initially appeared as a preprint, the sample was small, and only 18 people completed the fourth session. It measured EEG connectivity during a specific writing task, not physical brain atrophy. The results are nevertheless a serious signal. If the tool replaces the actual thinking process, less of one’s own process remains to remember later.

When AI checks AI

The National Autonomous University of Mexico provided an almost absurd example in 2026. For the first time, its entrance examination took place entirely online. A total of 158,712 applicants participated. The system used artificial intelligence, face and voice recognition, and human supervision to detect irregularities.

After unusual results, the university formed a technical commission. Eventually, 58,783 people were called in for an on-site verification exam. Media reports raised the suspicion that applicants had used AI and other aids outside the camera’s field of view.

Accuracy matters here too. The number 58,783 does not mean that all these people were proven to have cheated with ChatGPT. It denotes the group that was tested again in person. The case nevertheless illustrates the underlying problem: one AI monitors people who may be using another AI to circumvent the test. The machines check one another while humans become operators.

If an exam only measures who can use an aid without being noticed, it is not merely the exam that is broken. The learning that precedes it also loses its purpose.

How I use AI for this blog

I regularly use AI for research and corrections. I write the article’s first draft myself. During research, AI helps me find additional leads, work through long documents, or compare several sources. I then open and verify the decisive sources myself.

Once the draft is ready, I have it flag awkward word order, repetition, unclear transitions, and language errors. Sometimes the AI also points out a gap that requires more research. In this way, it mainly removes mechanical work and helps make a text I wrote myself cleaner and easier to understand.

Intelligence becomes a cheap commodity

The often quoted 20 dollars is only an order of magnitude: paid plans from various providers cost roughly that much, while capable models from Google, OpenAI, Anthropic, or Chinese providers are sometimes available for free. The important point is not the exact price, but that machine intelligence keeps becoming cheaper and more accessible.

For little money, systems can now produce texts, analyses, program code, and solutions to demanding mathematical problems. Leading systems already achieve gold-medal-level performance in mathematics and programming competitions, although not every model solves every task reliably.

This creates a dangerous temptation: why still learn to write, calculate, or program yourself? Because only genuine understanding helps us recognize incorrect results. The cheaper answers become, the more valuable good questions and source criticism become.

The economic side belongs in the discussion

It would be too easy simply to accuse young people of laziness. Many of the old promises genuinely work less well for them.

The Federal Reserve Bank of New York reported an unemployment rate of around 5.6 percent and an underemployment rate of 42 percent among recent college graduates in the second quarter of 2026. Here, underemployment means that a college graduate works in a job that typically does not require a degree.

At the same time, the Federal Reserve’s wealth statistics show an enormous generational divide. In the first quarter of 2026, Baby Boomers held around 51.6 percent of US household wealth. The Fed’s very broad “Millennials” category, which includes everyone born from 1981 onward and therefore Gen Z as well, held around 11.3 percent altogether.

That is no excuse to give up fractions, reading, or communication. But it helps explain why the traditional equation of a good education, hard work, and secure advancement feels less credible to many. People who see degrees carrying high costs, entry-level jobs disappearing, and wealth concentrated among older generations will seek other status signals and other shortcuts.

That is precisely why I find blanket insults aimed at a whole generation too cheap. The environment rewards brief attention, visible impact, and immediate reaction. Then we are surprised when young people become good at exactly those things.

So, is Gen Z the dumbest generation?

If “dumb” means being born biologically less intelligent, there is no evidence for it.

If instead we ask whether core abilities such as reading, mathematics, attention, and independent expression are less well trained among many young people, several sets of measurements reveal a serious problem. Gen Z is not the only generation affected. But it is the first to have spent its entire youth with smartphones, endless feeds, and then generative AI.

The answer therefore cannot be to ban AI and pretend that time can be turned back. I do not want to give up these tools myself. The answer must be to use them deliberately and intentionally refuse to outsource certain abilities.

Read a long text without jumping to a summary after three paragraphs. Write a first draft yourself. Try a calculation before asking for the solution. Ask a follow-up question in conversation. Open a source. Endure a mistake. Finish a thought.

That sounds unspectacular. Perhaps that is precisely what resistance looks like now.

AI can take a great deal of work off our hands. It just must not take away the part through which we learn to recognize good work in the first place.

Until next time,
Joe