Author: Hilman NurjamanEditor: Hosea ImmanuelLatumahina

Imagine you are opening WhatsApp, scrolling through a family group chat filled with three days worth of messages, and pressing “summarize”. In three seconds, dozens or even hundreds of messages about news, jokes, and grief are compressed into three lines of text. This is not fiction. Since Meta integrated AI summarization into WhatsApp and Instagram, millions of users now interact daily with a pre-processed version of their information, often without realizing it. The arrival of AI summarization in everyday platforms marks a shift far more fundamental than a mere feature update. This technology was once positioned as an optional convenience, a tool for those short on time. But when it becomes the default interface through which information is consumed, it no longer simply helps us read faster, it begins to define boundaries of what we consider “already read”. This is where a more fundamental problem begins.

We live amid what Hilbert & Lopez (2011) describe as an information pressure that collectively exceeds human cognitive capacity, every single day, the digital world produces content equivalent to billions of gigabytes. In this context, filtering information is no longer a luxury, it is a cognitive survival skill. AI summarization, in turn, opens doors to previously inaccessible knowledge from government policy documents to technically dense academic articles. A UNESCO report notes that such technology can help bridge limitations of time and language proficiency, broadening public access to collective knowledge. Yet behind this expansion of access and efficiency lies a set of consequences worth examining closely, regarding how we read, understand and manage information.

From Deep Reading to Surface Skimming

In her book Reader, Come Home (2018), cognitive neuroscientist Maryanne Wolf explains that deep reading (slow, reflective, and contemplative reading) is a process that neurologically trains critical thinking, empathy, and analogical reasoning. Reading, in this sense, is not merely the reception of information, it is a cognitive process that shapes how we reason.

Efficiency comes at a price. A similar warning was issued earlier by Nicholas Carr in his essay“Is Google Making Us Stupid?” published in The Atlantic (2008). Carr showed that internet search technology had already begun to shift how the brain processes text away from deep, linear reading and toward rapid, fragmented skimming. If search engines alone have altered how we think, then AI summarization which eliminates the act of reading altogether, carries the potential for far greater impact. More troublingly, AI summaries are never truly neutral. Every act of compression involves selection and omission, and what is frequently lost is precisely what matters most such as nuance, context, and argumentative complexity, the very elements essential to genuine understanding.

Feeling Informed Without Being So

The most dangerous consequence of this consumption pattern is the phenomenon psychologists call the illusion of knowledge. Rolf Reber & Rainer Greifeneder (2017) demonstrated that the ease of processing information—what they term processing fluency—is routinely misinterpreted by the brain as a sign of deep comprehension. When someone reads a neatly structured, smoothly flowing summary, the brain tends to signal “oh, I understand this”, when in reality, only pattern recognition has occurred, not genuine understanding.

The risk is compounded by the fact that AI summarization is not always accurate. LLM (Large Language Model) can produce text that appears coherent but is factually wrong. This phenomenon was called stochastic parrots, a condition in which language models mimic the patterns of text without truly comprehending the meaning behind it. In practice, such cases have already emerged in 2023, several AI-driven news platforms in the United States were reported to have published summaries that distorted the meaning of original articles, triggering public corrections from newsrooms. When flawed summaries are consumed by millions of users who never verify the source, the risk of disinformation ceases to be incidental, it becomes structural.

A New Form of Digital Divide

AI summarization also risks deepening existing inequalities, albeit in subtler ways. The digital divide has shifted from questions of device ownership toward what calls the skills divide, that gap in the ability to use technology critically and productively. In this context, users with higher digital literacy tend to treat summaries as a starting point for deeper exploration, while those with lower literacy are more likely to stop at the simplified version. The result is a paradox of democratization, technology designed to equalize access may instead widen the gap in comprehension.

At the same time, on the cognitive level, reliance on AI summarization risks fostering excessive cognitive offloading. That easy access to digital information changes how humans form and retain memory, the brain increasingly “delegates” the act of remembering to external technology. This aligns with Sweller (1988) cognitive load theory, which argues that the artificial reduction of cognitive effort can impede the formation of long-term knowledge schemas, precisely those needed for critical and analytical thinking. In short, the short-term convenience offered by AI may be paid for with the gradual erosion of long-term cognitive capacity.

These effects extend further into the socio-emotional dimension. Consuming information in fragmented, decontextualized form risks weakening the capacity for reflection and empathy. Reading longer texts trains tolerance for ambiguity and the ability to suspend judgement, two capacities essential for engaging with complex social issues. The Reuters Institute Report (2024) further recorded a growing tendency toward news avoidance among younger users accustomed to condensed content, suggesting that repeated exposure to instant formats can lower tolerance for informational complexity. In this regard, AI summarization does not only change how we read, but it gradually reshapes how we think and respond to the world around us.

Conclusion

In the end, AI summarization presents a dilemma we cannot ignore. On one hand, it eases our access to information, it alters how we understand that information in its entirety. This technology is not neutral, it actively shapes habits of thought, from how we read to how we determine that something has been “understood”. The challenge, therefore, is not to halt its use, but ensure we do not stop at the summary itself. Without critical awareness, the efficiency AI offers can quietly become a fragile illusion of understanding. In an age of ever-accelerating information flows, the ability to slow down, read more deeply, and think reflectively and contemplatively grows not less important, but more not as an outdated relic of the past, but as the very foundation of genuine comprehension, rather than the mere feeling of knowing.