The Rot of AI Writing
“This changes everything.”
There is an interesting thing happening online at the moment, and once you notice it, it becomes very hard to ignore.
People are getting tired of certain types of writing.
Not just specific words. Not just individual phrases. Whole sentence structures. Whole rhythms. Whole little copywriting patterns that we have all become used to seeing online.
The kind of language that used to feel clear, polished, and professional now often feels strangely empty.
You read something and, before you even know why, your brain flags it.
It feels too smooth. Too familiar. Too neatly packaged. Too much like it has been passed through the same machine as everything else.
And the annoying part is that a lot of this language was never bad in the first place.
Most of these phrases became popular because they worked. They helped people structure ideas quickly. They made writing easier to scan. They gave online posts a nice rhythm. They helped writers create tension, move between points, and make ideas feel more digestible.
Then AI came along and swallowed the whole style.
Now the same phrases that once made writing feel sharp often make it feel stale.
The phrases were already everywhere
I am not separate from this either.
I have used these phrases forever. Most people who write online have.
“Game changer.”
“Let’s unpack this.”
“The future is here.”
“Move the needle.”
“This changes everything.”
“Here’s the thing.”
“But there’s a catch.”
These were not originally AI phrases. They were internet phrases. Marketing phrases. LinkedIn phrases. Newsletter phrases. Copywriting phrases. They were part of the language of online attention.
For years, people were taught to write in a way that was short, clean, and easy to skim. You had to hook people quickly. You had to make the point obvious. You had to create little moments of tension so people kept reading.
So these phrases became shortcuts.
They told the reader, “A point is coming.”
They told the reader, “This is the important part.”
They told the reader, “Here is the lesson.”
And for a while, they were useful.
The problem is that AI learned the same shortcuts.
It was trained on the writing that already existed online, and that writing was full of these patterns. Blog posts used them. Business articles used them. Brand websites used them. Startup founders used them. Marketers used them. Thought leaders used them. Everyone used them.
So when AI learned how to write in a way that sounded professional or persuasive, it naturally learned the phrases that had already been rewarded across the internet.
Then people started using AI to create more writing.
And AI kept reaching for the same familiar patterns.
Then those patterns became even more common.
Then they became exhausting.
AI did not invent bland writing
This is the important part, (See what I did there).
AI did not invent bad writing. Humans were already producing plenty of bland writing before ChatGPT showed up.
Corporate copy was already full of vague promises. LinkedIn was already full of dramatic hooks. Marketing websites were already telling every company they could “unlock growth” and “streamline workflows” and “drive meaningful impact.”
AI did not create that world. It just scaled it. That is why the fatigue feels so intense.
We are not just reacting to AI. We are reacting to the mass production of a writing style that was already becoming tired. Before AI, this kind of writing was annoying but manageable. You saw it in sales emails, business blogs, motivational posts, and startup landing pages.
Now it is everywhere.
It is in emails.
It is in captions.
It is in articles.
It is in job posts.
It is in product pages.
It is in internal updates.
It is in people’s personal brands.
It is in comments.
It is in places where nobody even needed polished writing in the first place.
The internet has always had a lot of filler. AI has just made the filler faster, cheaper, and more confident.
Why AI loves these phrases
AI is not using these phrases because it thinks they are beautiful. It is not sitting there with taste, intention, or judgment. It is predicting what usually comes next. And in a huge amount of online writing, what usually comes next is a familiar copywriting move.
A dramatic opener. A clean contrast. A neat transition. A confident takeaway. A tidy final line.
These patterns appear over and over because they have worked for humans before. They performed well. They sounded professional. They made posts more shareable. They helped people package ideas.
So AI treats them as safe options.
If you ask AI to write something “clear,” “professional,” “engaging,” or “thought leadership style,” it will often reach for the most statistically familiar version of that request.
That means you get the same phrases again and again.
Not because AI is trying to be annoying.
Because it has learned that this is what polished internet writing looks like.
The problem is that what looks polished to the model now feels overused to the reader.
There is a gap between what the machine predicts and what the human brain wants to feel. The machine is optimising for fluency. The human is looking for life.
The em dash became a victim
The em dash is a good example of this.
People now point to em dashes as if they are an obvious sign of AI writing.
Which is funny, because human writers have used em dashes forever.
They are useful. They create rhythm. They let you add an aside without fully starting a new sentence. They can make writing feel more conversational or more literary, depending on how they are used.
The em dash did nothing wrong.
But AI started using it constantly.
It likely picked it up because em dashes often appear in polished writing. Essays use them. Articles use them. Opinion pieces use them. Marketing copy uses them. Writers use them when they want to sound thoughtful, fluid, or slightly elevated.
So AI learned that em dashes are associated with “good writing.”
Then it started putting them everywhere.
Now, when people see too many of them in a short piece, they become suspicious.
The punctuation itself is not the issue. The overuse is the issue.
That is what has happened to a lot of phrases too.
A phrase can be useful once. It can be annoying when repeated too much.
And it can become unbearable when a machine starts injecting it into every possible context.
The brain gets tired
The reason this matters is because reading is not just about understanding words. It is also about feeling texture.
When every piece of writing has the same structure, the brain starts to disengage. It begins to recognise the pattern before it has processed the thought.
You see the hook coming. You see the contrast coming. You see the “takeaway” coming. You see the neat little closing line coming.
At that point, you are not really reading the idea anymore. You are reading the template. That is where the fatigue comes from.
It is not just that these phrases are overused. It is that they make the writing feel pre-decided. They make it feel like the conclusion existed before the thinking did.
And that is boring. Even when the content is technically correct, it can feel dead.
There might be a point in there somewhere, but it is buried under a layer of internet polish. The sentence is readable. The structure is clean. The grammar is fine. But the whole thing feels like it has no pulse.
The internet has developed an accent
There is now a kind of internet accent in writing.
You can hear it.
It starts with a punchy opening line. Then it gives you a short setup. Then it says something like, “But here’s the thing.” Then it adds contrast. Then it lands on a lesson. Then it ends with a sentence that sounds like it was designed to be screenshotted.
That structure is not always bad. Sometimes it is genuinely useful. There is nothing wrong with making writing clear or easy to follow.
The issue is that the accent has become too common. It is like stock music in YouTube videos. At first, you do not notice it. Then someone points it out. Then suddenly you hear it everywhere.
Once the pattern becomes visible, the effect is gone. The writing stops feeling like a person trying to communicate an idea. It starts feeling like a format trying to complete itself. That is what a lot of AI-assisted writing now feels like.
Not wrong. Not unreadable. Just strangely pre-chewed.
Writing differently is not a perfect escape
The obvious answer is to write differently.
Avoid the phrases. Break the patterns. Use stranger examples. Be more specific. Keep more of your actual voice in the piece.
That is all good advice. But there is a slightly cursed problem here.
As soon as enough people start writing differently, AI learns that too.
If everyone moves away from polished LinkedIn language and starts writing in a rougher, more casual, more obviously human style, the machine will absorb that style as well.
Then we will start seeing fake roughness.
Fake casualness. Fake personality. Fake typos. Fake “I just had this thought while walking my dog” energy.
That is the trap.
Style alone is not a permanent escape. AI can copy style. It can copy rhythm. It can copy roughness. It can copy the appearance of personality.
What is harder to copy is actual thinking.
Specific experience.
Real taste. A strange opinion. A personal frustration. A detail that could only come from someone who was actually there.
That is where the difference still lives.
The problem is not using AI
I use AI all the time.
Mostly for editing.
This article is a good example of how I think AI can be useful.
The raw version of this started as a messy thought. Basically a brain dump. It had typos, half-finished points, missing words, and the usual chaos that happens when the idea is moving faster than the typing.
That is normal.
A lot of good writing starts messy.
AI is helpful when it acts as an editor. It can help structure the mess. It can make sections clearer. It can smooth out a paragraph that is confusing. It can help turn a raw thought into something another person can actually read without having to decode your brain.
That is a win.
The thought is still yours. The angle is still yours. The frustration is still yours. The lived experience is still yours.
AI is just helping clean up the room.
The danger is when people skip the thinking part and ask AI to invent the whole thing from nothing.
That is when the writing starts to rot.
Empty prompts create empty writing
When you ask AI something broad like, “Write me an article about AI in business,” you are probably going to get a broad, generic answer.
That is not surprising.
There is no real point of view in the prompt. No specific experience. No tension. No argument. No personal observation. No reason for the article to exist.
So the model fills the space with the most available material.
And the most available material is generic internet language.
You get “AI is transforming the way businesses operate.”
You get “companies must adapt to stay competitive.”
You get “the future belongs to those who embrace innovation.”
You get all the usual filler.
Again, it is not technically wrong.
That is what makes it worse.
Bad writing that is obviously wrong is easy to reject.
Generic writing that is technically fine is harder to fight because it passes a basic quality check. It has a headline. It has sections. It has transitions. It has grammar. It has a conclusion.
But it does not really say anything.
It just performs the shape of saying something.
The junk drawer of thought leadership
This is why so much AI-generated business content feels empty.
It has all the signals of intelligence without much actual intelligence inside it.
There is structure.
There is confidence.
There is vocabulary.
There is a polished cadence.
But the piece often feels like it was assembled from the junk drawer of online business language.
A little bit of strategy.
A little bit of transformation.
A little bit of innovation.
A little bit of “fast-paced landscape.”
A little bit of “unlocking potential.”
A little bit of “the bottom line.”
And there you go.
You have an article.
Sort of.
The issue is that readers are getting better at sensing this. They may not be able to explain exactly why something feels AI-written, but they can feel the absence of a real person.
They can feel when the writing has no stakes. They can feel when the author is not risking an opinion. They can feel when there is no actual thought underneath the polish.
Some normal writing will get unfairly punished
The sad part is that a lot of normal human writing is going to get caught in the blast radius.
A human can use the phrase “game changer” sincerely. A human can write “the takeaway” because sometimes there really is a takeaway. A human can use an em dash because it fits the rhythm of the sentence.
Not every familiar phrase means something was generated by AI.
But the context has changed. Readers are more suspicious now.
Certain phrases carry baggage they did not carry before. They might still be useful, but they arrive with a little static around them.
That is what AI has changed. It has not made these words illegal. It has made them tired. And tired language is dangerous because it makes readers switch off before they have even reached the point.
The real fix is not avoiding every phrase
I do not think the answer is to make a giant banned list. That just creates a new kind of weird writing. You end up with people avoiding normal language so aggressively that the writing becomes awkward in a different way.
The goal is not to never use a familiar phrase again.
The goal is to know when you are using one because it is the right phrase, and when you are using one because you have stopped thinking.
That is the difference. Sometimes “the bottom line” is fine. Sometimes “here’s the thing” is fine. Sometimes a clean transition is useful. But if a piece is held together entirely by those phrases, there is probably not enough actual thought in it.
Write the messy version first
The better process is probably simple. Write the messy version first.
Get the actual thought out before you start polishing it. Do not begin with the template. Do not begin with the hook. Do not begin by asking AI to sound clever on your behalf.
Start with what you actually think. Start with the frustration. Start with the example. Start with the thing you noticed that made you want to write in the first place.
Then use AI after that if it helps.
Use it to organise. Use it to clean. Use it to make the argument easier to follow. Use it to find the section that is confusing. Use it to turn the raw material into something more readable. But do not let it replace the raw material.
That is the part people actually want.
AI can polish, but it cannot care
This is probably the real line.
AI can polish. It can structure. It can imitate. It can produce competent sentences forever.
But it does not care.
It does not get annoyed by overused phrases. It does not feel the boredom of reading the same hook for the thousandth time. It does not have taste in the human sense. It does not have that little internal reaction where you read a sentence and think, “Please, not this again.”
Humans do.
That irritation is a form of taste. And taste is becoming more important now, not less. Because when everyone has access to the same tools, the differentiator is not who can produce the most polished paragraph. The differentiator is who can tell when the polished paragraph is actually boring.
The copywriting shortcuts are cooked
So yes, the phrase fatigue is real.
The rot is real.
A lot of online writing now has a weird sameness to it, and AI has made that sameness much harder to escape.
But I do not think the answer is to panic every time we see an em dash or a familiar phrase.
The answer is to stop using these things as substitutes for thought.
A strong idea can survive a simple phrase.
A weak idea cannot be saved by a polished one.
That is the part worth remembering.
AI is useful when there is something real underneath it.
It is painful when there is not.
Use it to edit.
Use it to structure.
Use it to clean up the mess.
But make the mess yourself.
End of the day, write better.