AI slop has been a mild obsession of mine since sometime in 2024. I’ve finally decided to take some time to write a blog post about it. By blog post, I mostly mean rant.
If you use AI for writing, you’ve probably had this experience. You ask for an email, a summary, a LinkedIn post, whatever. The result is perfectly competent. Nothing is obviously wrong with it. And yet, somewhere around the second paragraph, you think, “Yep. ChatGPT wrote that.”
Sometimes you can point to the problem. Often you can’t. It just doesn’t sound quite like a person. There’s a reason for that. AI does not understand language the way we do. If I text you, “Call me now.” that means something very different depending on whether I’m your boss, your mother, or someone you met at a bar last night.
Change it to “Call me now…” and suddenly we may have an entirely different situation on our hands.
Humans understand this almost instantly. We know who is talking, what happened before, whether they’re pissed off, flirting, panicking, being sarcastic, or just terrible at punctuation. Most of the time, we don’t even realize how much information we are bringing to those three little words.
AI has to figure all of that out from clues. Sometimes it does a surprisingly good job. Sometimes it writes something that sounds like the communications department of a regional bank.
That’s where synthetic residue comes in. Synthetic residue is the term we use at HIA for the patterns AI tends to leave behind in writing. In plain terms, it’s the stuff that makes you read something and think, “Yep. ChatGPT wrote that.”
Em Dash, Thou Art Mine Enemy?
Everyone knows about the em dash by now, so I won’t spend much time on it. I will simply note that I have managed to write this entire section without one and accept your congratulations. The em dash is only the most obvious example.
Patterns Are the Main Driver
One of AI’s biggest patterns is… patterns. AI learns from enormous amounts of language and gets very good at predicting what usually comes next. That ability is a large part of why it can write so well in the first place. It is also one of the reasons its writing can become so boring.
AI likes familiar structures because familiar structures are statistically safe. Neat setup. Explanation. Example. Summary. Then another setup, explanation, example and summary. If you have ever read something that was clear, correct and somehow completely forgettable, you know the feeling.
Human writing is usually messier. We linger on the part we care about. We rush through the obvious part. We interrupt ourselves. We spend an entire paragraph on something another writer might dismiss in one sentence. That irregularity is not necessarily bad writing. Sometimes it is the thing that tells you there is actually a person behind it. AI often needs help preserving that.
Variety Is the Spice of AI Life
AI also loves variety. Especially adjectives. Give it one perfectly good word and it will spend the next six paragraphs desperately avoiding using that word again. Suddenly “useful” becomes practical, valuable, impactful, effective, meaningful and transformative. Relax, Claude. Useful was fine.
This gets more confusing when AI starts changing nouns. A process becomes a framework, then an approach, then a methodology. Now your reader has to stop and wonder whether those are actually different things.
As writers, we sometimes assume repetition is automatically bad. Sometimes it’s perfect. If you are talking about the same thing, using the same word can be the clearest possible choice. Nobody gets a prize because you found seven synonyms for “process.”
It Loses the Plot, or Misses It Entirely
Some of AI’s writing habits are mildly annoying. Others can actually change the meaning of what you are trying to communicate.
Upload meeting notes without good instructions, for example, and AI may assume the thing mentioned most often was the most important thing. That sounds reasonable until you remember how meetings actually work. Someone senior says something once. John likes it. John repeats it five times because John would very much like the senior person to notice him and, ideally, promote him. AI sees a recurring theme. Everyone else sees John doing the John thing.
Frequency is not importance. Discussion is not a decision. And repeating something loudly does not make it smart.
The people in the room understand things that may never appear in the transcript. They know who has authority, which ideas went nowhere, what was decided, what was merely discussed, and when somebody was talking because they had something important to add versus because they enjoy hearing themselves talk. AI may not know any of that unless we tell it.
We Humans Are Part of the Problem
This is the point where it becomes tempting to blame the machine. I know because I do it too. When AI gives me something bland, repetitive or strangely corporate, my first instinct is often to think, “Why did you write it like that?” Fair question.
There is another question that is probably more useful: “Why was I about to send it?” I could keep going through the patterns. We teach them throughout our foundational HIA courses because once you start noticing synthetic residue, you see it everywhere.
But AI is only half the problem. The other half is what we do with the draft after AI gives it to us. Most of us started using AI because we wanted to save time. That is perfectly reasonable. If a tool can help you write an email in three minutes instead of twenty, there is no virtue in spending the extra seventeen minutes staring nobly at a blank screen.
The trouble starts when “AI saved me time writing this” becomes “therefore I don’t need to read it.” And I have caught myself doing versions of this too. You ask AI to write something. It looks pretty good at a glance. You are busy. The next task is waiting. Copy. Paste. Send. Then somebody else has to read it.
The strange thing is that before AI, almost nobody believed the first draft of something important should automatically be the final draft. We reread our work. We cut sentences. We realized we had made the same point twice. We changed a word because it didn’t sound like us. AI made the first draft dramatically easier. It did not make the rest of writing unnecessary.
The HIA Golden AI Writing Rule
The golden rule we teach at HIA is simple: estimate how long it will take someone to read what you created with AI. Spend at least that long editing it. If it takes your reader five minutes to read, spend five minutes reviewing it. That seems harmless until you look at the equation from the other side. If twenty people are going to spend five minutes reading something you created, you are asking for 100 minutes of other people’s attention. Five minutes of yours starts to look like a pretty reasonable investment.
Notice I said editing. Not prompting. Good prompts matter. Better instructions usually produce better drafts. But we have developed a peculiar habit of spending fifteen minutes trying to persuade AI to fix a sentence that we could fix ourselves in thirty seconds.
At some point, put down the prompt and become the editor. AI can still help. Ask it where the writing gets repetitive. Ask which parts are confusing or bloated. Ask whether it keeps changing terminology. Ask what sounds generic. Ask it to point out patterns you may have stopped noticing because you have been staring at the document for an hour.
Then read it yourself. Your name is still on it. That may be the part of AI writing we need to remember most. The machine can create the draft. It can suggest improvements. It can catch things we missed. We still have a reputation to protect.
Happy writing.