Wrong, writers aren't being replaced by AI. Readers are.

Who are we writing for, actually?

A Reading From Homer, Lawrence Alma-Tadema

When ChatGPT was first released by OpenAI in 2022, everyone and their mothers seemed to be so sure that it would one day replace writers. After all, it’s able to do a surprising number of things well, including writing something that seems convincingly human, so it doesn’t seem like this prediction is too far-fetched.

In fact, last year, Microsoft released a study detailing job roles that are most at risk of and most protected from AI disruption. Though it’s worth noting that the study does not say that AI will take away jobs. Rather, it shows that AI, particularly LLMs, will change how work is done by automating some tasks, leading jobs to adapt instead of vanish entirely.

True to everyone’s prediction, writers and authors are ranked in the top 5 of roles that are most at risk.

So, how are things now, around 4 years after Pandora’s box was opened?

Fortunately, I still write for a living, and I don’t mean as published author (yet), but a regular marketing job that’s been haunted by the curse of automation for the past four years. But things have changed. Significantly.

The amount of time I usually spent on research has been cut down drastically. Some days, my tasks revolve more around heavily refining a mediocre AI-generated draft rather than writing from scratch. The demand for writers has declined significantly. I’ve seen more and more text posts on social media that show obvious signs of AI writing.

Kind of bleak, yes, if not extremely bleak. But as a writer, there is something I find more alarming than the impending death of writers. It is the fact that readers are the ones being replaced.

We don’t really read anymore

Reading, Berthe Morisot

By “reading,” I don’t mean the mere consumption of text. We are surrounded by words more than ever. What I mean is reading as an active process: slowing down, holding ideas together, interpreting meaning, and engaging critically with what’s in front of us.

For the longest time, reading has been the primary way to gain knowledge and information. Even with the internet, we still needed to read — or, at least skim an article — to get the insights and answers we were looking for.

Then, short-form video content became popular. Initially, it was nothing more than harmless entertainment, popularized by platforms like Vine and Snapchat. Then TikTok was created and it quickly had the rest of the world in a chokehold, even prompting other social media platforms to replicate its format (e.g., Instagram Reels and YouTube Shorts).

By this time, short-form video content had evolved from a source of entertainment into… pretty much everything. A product review, a cooking tutorial, a quick mental health ‘diagnosis’, skincare tips, a branding insight.

Every piece of information is delivered in a video that is quick, simple, and concise. A video that you can watch or listen to while you do your own routine. In fact, a 2024 study by Adobe found that TikTok is the fourth most popular search engine.

While short-form videos train us to expect answers that are instant and simplified, LLMs take it to the next level. Like short-form content, AI can give us direct answers without requiring us to browse through pages of books or multiple websites. Even if they don’t necessarily use ChatGPT as a search engine, Google has taken the liberty to present a comprehensive AI-generated overview at the top of its search results.

The interactive nature of LLMs also makes them more appealing than Google. Ask ChatGPT to help you brainstorm a thesis idea, and it’ll easily give you links to several resources and even entertain questions like, “I don’t understand this paper, can you tell me on which page I’ll find the data I need?”

Of course, Google, as a search engine, is not without its faults. Sometimes the algorithm just sucks. Many times, it gives us 10 different variations of the same articles that don’t even give us the answer we need (thanks, SEO), or worse, pure clickbait. So, maybe it’s not all that bad?

But a 2025 study comparing the effects of LLMs versus web search found that when search becomes faster and easier, people read less, compare fewer sources, and think less on their own. As a result, learning becomes shallower because people engage less actively with information.

AI is changing the way we write… for the worse

Portrait of a Young Lady, Master of the Female Half-Lengths

The way Google search works is through SEO, or Search Engine Optimization, which determines which pages appear when we search for something. For the longest time, SEO has relied on two important elements: good technical skills and good writing skills.

In fact, I think it won’t be too bold to say that in the 2000s and 2010s, there would have been no search results without writers, because search engines relied entirely on human-written content to function.

Back then, writers focused on writing pages that answered real questions, using languages that humans actually read and searched for. Writers write for humans. Every topic they researched, every outline they made, every word they chose, was done with the intention of making human readers understand whatever it is they’re searching for.

Now? Not so much. Nothing is meant to stay the same, after all. SEO has shifted to Semantic SEO, where content is evaluated by meaning, context, and how ideas relate to one another. The reason is that LLMs process content semantically. They don’t look for exact phrases or keywords, but evaluates whether the content is coherent, informative, and semantically complete.

This technicality might sound boring, or if you’re still reading, you might think this is a good thing. In some way, I suppose it is. But the thing is, as more and more people rely on AI-generated answers without further exploration, content visibility itself no longer depends on ranking as a clickable link, but on being included in AI summaries and overviews.

Semantic SEO is a system that is rather rigid for human writers. It prioritizes clarity, speed, and predictability. Answers are expected to be delivered quickly, ideas structured neatly, and stylistic experimentation deprioritized. All of this serves one goal: producing short, digestible chunks of information that can be easily extracted by AI.

Who are we writing for, again?

Pandora, John William Waterhouse

If the words from the last paragraph haven’t truly sunk in, I’d like to repeat them again: Written content is now made to be as short, simple, and predictable as possible so it can be easily extracted, summarized, or reused by AI systems. There are so many implications of this shift that I honestly don’t even know where to start and how to write them in the right order.

To start with, when content is made for the sole purpose of being fodder to a machine, then everything would sound the same. Ever seen the AI Overview on Google? Well, imagine if the majority of content on the internet is written in the same way, the same tone, the same format. The same random bolded text, the same overuse of em dash, or formulaic phrases like ‘It’s not A. It’s B.’

But then again, perhaps this doesn’t really matter because people barely read things on the internet anyway. So perhaps, for the most part, creativity and writing style don’t really matter. Maybe it doesn’t really matter how something is written, because what’s more important is whether the content fits what AI considers a ‘good answer’.

If it’s not good, the content would probably sit on the seventh or fifteenth page of Google, which probably doesn’t make any difference now, considering many people barely scroll past the AI Overview. If it’s good, then the content would be summarized and regurgitated by AI in whichever tone they decided to use for the day.

Writing is an act of communication between humans, but now it has mostly been reduced to raw materials, data input, or infrastructure. So then, what happens to explanation, or nuance, or persuasion? What happens to writing that only makes sense when read slowly?

The death of the readers

Cassandra, Evelyn de Morgan

Perhaps it is simply human nature to favor things that are quick and easy. I do admit it is extremely hard to resist the urge to skip several steps so we can get straight to the end. But there is value in process, in slowness, and in taking time.

Reading itself is not a passive activity. It trains our attention, memory, and critical thinking. Long-form reading requires us to hold ideas in our heads, connect arguments, and tolerate ambiguity. So when information is delivered instantly without complexity, people engage less actively with ideas. In fact, a 2025 study by the MIT Media Lab indicated that heavy reliance on AI can be linked to reduced cognitive engagement. In other words, people remember less, synthesize less, and rely more on external systems to do the thinking for them.

This isn’t something that will happen in the future. We can already observe it in how the media is produced and consumed today. For example, movies are now becoming simpler due to humans’ declining attention span.

Culturally, this also narrows what writing can be. Texts that demand time, slowness, or interpretation become increasingly rare because they don’t perform well in a system optimized for speed and extraction. Writing still exists, but often without being read. And when reading disappears as a habit, so does the mental discipline it cultivates.

What we lose, then, isn’t just literature or authorship, but the conditions that once allowed people to think slowly and deeply on their own.