A while ago I wrote a piece about the importance of genuine writing in the age of ChatGPT, where I advocated that writing should be preserved and approached as a process-driven activity, not something to be always optimized for. However, since then things have changed a bit. Although this still holds true for the present essay, I have to admit something that I consider a flaw and which I feel guilty and uncomfortable about:
With the excuse to meet productivity goals, increasingly I have surrendered myself to the conveniences offered by LLMs. The goal in mind is obvious: to expedite my thinking and crafting process (as if that was possible anyway).
I’ve been noticing how this habit has been negatively impacting my relationship to the writing process when I want to do it for reasons other than work, such as writing the present essay or a hand-written letter to a friend for a special occasion.
Every time I use LLMs for writing, I remind myself that the beauty of writing lies precisely in the awfulness of having to navigate an unclear path to declutter my thoughts. But LLMs have transformed what we once understood by writing. One can argue that typewriters also transformed the meaning and experience of writing. Our calligraphy got worse, but we can type words much faster than handwriting. The need for speed and increased flows of communication is what made typewriting dominant in our world. Because we are in general more concerned with speed over quality, LLMs fit us like a glove.
I am a victim of this collective obsession, and that’s why I generally use LLMs: out of convenience. Perhaps you might have similar struggles to mine. If so, the following lines are an attempt to unfold why, at a personal level, that is the case. Then, I try to provide a way-forward on how to overcome it.
First, I began to be aware of how often I compare my writing style and ways I structure my essays to how the LLM organizes information and arguments, which normally leads to me thinking that LLMs do a better job at crafting, editing and polishing my text in ways that I wouldn’t be capable of myself. This led me thinking that my writing should be more machine-looking: clean, objective, linear reasoning, with extremely structured paragraphs and outline, rather than trying to explore my own style based on human authors.
Secondly, I have increasingly become more impatient to commit to the slow, messy, unclear and non-linear process of writing. Because LLMs are only one-click away, I can notice that my cognitive endurance to focus on developing my own ideas and writing structure is reduced. I feel it’s too tempting to avoid using it.
Another day, the Instagram algorithm— knowing I was researching on the topic —assertively brought a cut from an interview with Akira Kurosawa, a Japanese screenwriter and film director, in which he talks about the dull task of writing. As an old-school kind of person that he was, Kurosawa said quite simply that “the most essential thing a writer should have is the patience to write one word at a time and not rush to the final destination.” He illustrates by using an analogy between writing and mountain climbing: “if you want to climb a mountain, the first thing you’re told is not to look at the peak, but to keep your eyes on the ground as you climb. If you keep looking at the top, you’ll get frustrated.”
Kurosawa’s words have deeply resonated with how I perceived that LLMs impacted my approach to writing. It’s very true that one of the main challenges for writers is to write the first line. How to start? Which words to pick? How to translate all the chaotic ideas into a steady, linear, structured way? These are the common questions writers constantly struggle with.
Perhaps Kurosawa’s analogy didn’t account for a significant difference between mountain climbers and writers. Whereas the first climb towards the mountain’s peak is clear—with maybe only partial detours—the writers’ path is not unidirectional, and many times the ‘finishing line’ is not known from the start. A writer goes back and forth in his texts several times (this essay being an example), attempting to articulate sentence by sentence, but he always and at the same time is concerned with the understanding of the whole. It’s a double-edged endeavour in philosophy understood as the hermeneutic circle of understanding.
In our society obsessed with outputs, LLMs give us the possibility to clear this messy path, making it easier for writers to jumpstart and arrive at the final text (or a draft) much quicker.
Given my background in Philosophy of Technology, I tend to be very aware and critical about my approach to the tools I use, how I use them and how they shape my ways of living, working and thinking. In that sense, my relationship to LLMs is a love-hate one. The love side is because LLMs are undeniably a great tool for many purposes. The hate side is because LLMs are undeniably a great tool for many purposes.
Caught in this love–hate relationship, I find myself in a continual process of negotiation. Both the denial of technology and an uncritical embrace of it foreclose the possibility of a free relationship. Over time, I have realized that my writing takes shape in two distinct approaches: writing as a productive activity and writing as craft.
Writing as a productive activity does not always involve personal motivation. Seeing it this way, writing becomes a means-to-an-end, and therefore, it is output-oriented. If you need to write, for example, a report linking data points for a corporate presentation, chances are that your willingness to write everything from scratch is not motivating enough. Therefore, you immediately turn to LLMs so that they can help you out with the heavy lifting, sparing some energy to your brain and make you sound more ‘productive’. We soon realise how convenient LLMs are to perform tasks quicker.
This is even more the case when we write in a foreign language. Even after more than a decade of working in English, there remain parts of my life that I can communicate only in Portuguese.
Not by coincidence, the philosopher Martin Heidegger said that ‘language is the house of the being’. Language is much more than an orderly combination of letters and words. We feel, think and express ourselves through language. We do not use words to communicate meaning, we inhabit them.
LLMs offer a way to strengthen my English writing almost instantly. The effect resembles that of a tourist repeatedly visiting a city: fluent enough to navigate, yet never compelled to truly settle. This convenience makes me lazier, less willing to invest the sustained effort required to perfect my writing, where ‘perfection’ is understood here not through benchmarks measuring quality but existentially, as the act of becoming at home in a non-native language. In this way, LLMs may be turning us all into tourists of English, very frequently visiting the country. However, at the same time, thanks to writing with LLMs, many more might develop this tourist-like level of writing in multiple non-native languages, which would be non-realistic without the existence of these tools (or a lot of talent and discipline).
However, even if we might agree that it’s okay to use LLMs to accomplish work in less time, and even if we acknowledge it can be helpful to improve my writing capabilities in multiple languages, from a cognitive standpoint, we should not overlook the side effects of this process.
Writing requires practice. And it’s a lonely, time-consuming exercise, which obliges us to think a lot. The more we get used to outsourcing this process to LLMs (i.e. cognitive offloading), the less we become inclined to engage in this process at all. As the tools get better, the more we are inclined to trust that they will deliver the expected outcome, with less required effort from our side.
There are many empirical studies that try to explain why we are prone to offload thinking and memory. Among the various reasons, one often cited is that our brains have a finite energy-capacity bandwidth. As such, cognitive offloading allows us to free up mental energy and concentrate on other tasks we find more valuable pursuing. For instance, if writing is primarily interpreted as a means-to-an-end task (let’s say, for an essay exam or a work report), chances are that we will be more inclined to minimize the struggle and effort of writing.
This results in some serious side effects. A 2025 study led by MIT researchers measured the cognitive impacts of using LLMs for writing against those who don’t. Participants were split into three groups: ‘LLM users’, ‘search engine users’, and ‘brain-only users’, which were tasked to write an essay. LLM users displayed significantly less neural activity in the regions associated with memory and critical reasoning in comparison to the other two groups. They also self-reported to have less sense of ownership of their essays and struggled the most to quote anything from their work after the writing was finished. Furthermore, researchers continued to monitor the cognitive impacts on the ‘LLM users’ for a period of four months and found that they consistently underperformed at neural, linguistic and behavioural levels.
The long-term cognitive effects of this are yet to be found, but they do not seem promising. Just like any muscle, our brains too face a risk of ‘atrophying’ if we don’t train them. Not by coincidence, the Oxford word of the year in 2024 was ‘brain rot’, which is defined as: “the supposed deterioration of a person’s mental or intellectual state, especially viewed as the result of overconsumption of material (now particularly online content) considered to be trivial or unchallenging. Also: something characterized as likely to lead to such deterioration”.
The equation seems to be inversely proportionate: the more we rely on LLMs, the worse our brains get. In a scenario like this, what can we do?
Craftsmanship was once ‘mainstream’—back when mainstream meant knowing how to make things yourself. Before industrial societies, automation, and the large-scale alienation of work, crafting implied a close relationship between subject and object: the person who made something had to know every step required to bring it into being. Because a craftsman cannot fully standardize the process, every crafted object is necessarily unique. Each one carries the unmistakable personal trace of its maker. Therefore, they cannot be mass-produced.
On the contrary, words spelled out by LLMs do not originate in meaning in a human sense. LLMs reduce language to a computational and standardised process, and in doing so subtly reshape how we, as humans, come to understand and relate to language itself. To use Heidegger’s conceptualization of modern technology, LLMs ‘challenge forth’ words: they are extracted from a vast online repository of data, transforming writing into a statistical, algorithmic operation. Language is commodified.
Writing as craft de-commodifies language by making it messy, slow and not easy to scale up as it is fully dependent on my labour. Choosing to fully embrace it in the age of LLMs becomes both an act of resistance against the continuous ‘challenging forth’ of words and a personal affirmation of our poiesis—our ability to cultivate existential meaning through the ‘bringing forth’ of words. Writers, as craftsmen, are merely the vehicle through which language unfolds.
Besides the short-term conveniences of cognitive offloading, perhaps one of the main reasons that LLMs are so tempting to use lies on the macro-level: LLMs have been developed with the promise to maximize our productivity exponentially, so we can save more time to engage in the activities that are more meaningful for us to pursue. And we are ingrained in this mentality.
The problem with this idea is that the more we get used to speeding up and automating tasks, the faster we become impatient with activities that require friction, such as writing or reading a challenging text. Therefore, it becomes increasingly hard for us to separate writing as a productive activity (LLMs accepted) and writing as craft (LLMs not desired).
From that, another question follows: is the pursuit of friction even desirable? I personally don’t think there is a one-size-fits-all type of answer. What I can say is that by mindlessly removing friction in our lives, there is a risk we set us further away from having the feeling of real accomplishment, something that feels genuinely ours, that has our trace, our touch, our habit.
That’s why I feel somehow uncomfortable when I use LLMs for writing, even for work purposes. I believe this discomfort comes because it feels like I’m cheating on the very activity of writing, that is, I am doing something else rather than writing, alienated instead of home. Rather, I am closer to a word-puzzle-aggregator, just conveniently editing and throwing new ideas to be algorithmically added to a text that has not been originally or mainly written by me. This is an entirely different process than an editorial curatorship, with the back-and-forth of suggestions, ideas and edits that come from another human being.
Yes, it takes a lot of time and effort - both ways. But in the end, there is ownership and a meaningful space we inhabit and dwell in.
Maybe it’s writing – not LLMs – too good to be avoided.
