1-LLMs could be a threat to more than our jobs. (Intro)
Nearly six years after LLMs became part of our modern daily lives, with millions of visitors (1) each day, it is undeniable that the appearance of LLMs had a deep and lasting impact on the way people browse the Internet, how they search and how they interact with it. This radical shift in behaviour has many consequences, like the increase of zero-click and AI overviews. One aspect of heavy LLM use that is of interest to me, though, isn’t zero-click navigation and the slow choking death of websites with plummeting ad revenues. My interest lies with linguistics. Words.
Much like social media in their time, I suspect LLMs' consequences (good or bad) will become visible progressively, slowly, and potentially subtly. It is probably still too early to know what the actual consequences of heavy LLM use on speech and writing will be at a societal scale, yet this is something that interests me.
We humans learn from each other constantly, including speech and words. We mimic each other, starting with our parents when we were toddlers. It is not too uncommon to pick up a new word or a new pattern of words after repeated exposure from a colleague or friend, or maybe even a show or book. This series of articles will explore my worries about massive LLM use as we see today, for so many people simultaneously, all being exposed to an identical pattern.
What if repeated and prolonged use of LLMs had a lasting impact on how we collectively write, speak and think? What happens when most of our Internet browsing is done through a chatbot, over and over again? I am worried that prolonged generalized use of LLMs could lead to a shift in the way we write, impoverishing our languages and leading to poorer communication. We could observe a homogenization of writing style and structure, potentially reinforcing preexisting issues linked to the reduction of the time people spend reading books. (2).
Because LLMs are owned by corporations, we do not know precisely what parameters and restrictions those companies choose to apply to their models. But we know they have the freedom to change the way models behave at will, without our agreement. If prolonged use of LLMs does end up having an impact on our writing and vocabulary, this puts the population in a unique position, a risky one. This would give those corporations the ability to curate words, and thus concepts. This could become an insidious form of propaganda or censorship. While I do focus mostly on the removal and loss of words, it is also true that they could push new words or, more likely, serve a word or phrasing that benefits them more frequently.
There is probably a case to be made that this is simply another step in the direction of a society where people will, possibly, own nothing. Not even their words. But why does it matter? Well. We communicate among ourselves using words, which reflect concepts. If I write “apple”, you understand the meaning of that word and usually picture an example of it in your mind. By writing a series of letters, I caused your mind to think of one apple. But words fade away, especially words expressing complex concepts like, for instance, “Hohmann transfer orbit”. If you have never heard of this concept, those words do not evoke anything to you. But to anyone working in the right industry or anyone interested in orbital mechanics, you grasp or fully understand the idea. But if you learn this concept and subsequently do not encounter or use it anymore for a decade, the superficial definition might still be known, but it is likely that the finer details will be lost to time and forgotten.
I suppose the same would happen to words erased or banned from LLMs. The word “liberty” could be taken away, replaced arbitrarily by “leeway” or “license” depending on context. Liberty carries with it a sense of inherent freedom, something you own. Leeway evokes something granted, temporary. I suspect total removal isn’t even necessary to see the effects and that simply favoring some words instead of the banned one would yield effects eventually in the population. But what happens in a society when the word “liberty” dies in favor of an arguably much weaker word for the individual like “leeway”?
Would we still have people trying to conceive a utopia if we stopped being exposed to the word utopia nearly entirely? I wager simply reducing the frequency of a word would be enough to limit or stop the generation of ideas around its concept nearly entirely.
If this idea of mine is true, how long before long lasting damage to our civilization are made? How long before we run out of thinkers who are free to imagine creatively, using any word or concept they want and how long until they are replaced by people thinking only within the confines of a curated list of words and concepts?
Liked what you read? You can support my work with an espresso on Ko-fi
Bibliography
LLM Statistics 2026: Where 800M Users Are Searching Instead of Google Evolv Agency. https://evolvagency.io/learn/generative-search/llm-statistics-2026.
Teens today spend more time on digital media, less time reading. https://www.apa.org https://www.apa.org/news/press/releases/2018/08/teenagers-read-book.