Help! AI Thinks I'm AI.

Yesterday, while in bed with my laptop on my belly, I browsed HackerNews.

One post caught my attention: LLM Writing Tropes.md 👀
I write online. I hate AI slop. TELL. ME. MORE.

So, I visited the the site and my thought process went something like this:

1. That's cool! Someone created a markdown file that you can use to avoid sounding like an LLM.
2. Wow, that's quite a collection of LLM tropes (i.e. stylistic choices that sound "robotic")
3. Wait a minute. The creator didn't write the markdown file himself. The disclaimer on the site says: "Disclaimer: Creation of this file was AI-assisted. If you thought I was going to write out a .md file for AI myself you must be mad. AI for AI. Human for Human." LOL
4. Oooh... there's a [VETTER] button. I bet you can check URLs and see if those are created by AI.
5. I copied my first climate fiction story (that I'm quite proud of), pasted it, and hit [Vet]

I wonder what it thinks of my writing 🤔

You know what was not on my 2026 bingo card: Being found guilty of tricolon abuse.

The thing is, I published this particular piece in June 2022, months before ChatGPT was made available to the public (let alone was capable enough to write a coherent story).

Here's my short story. Here's a look behind the scenes of creating it. Judge for yourself.

So what happened? Why does our AI buddy think I'm its colleague? Well, I have a guess.

To understand that, I'll have to take you back to my early online writing days.

Ground Zero of Sounding Like AI

I'm not a native English speaker. 

I learned the language in high school but improved it significantly during my professional career. First, talking to colleagues in English. Then, working in customer-facing roles in the US. And finally, building my own company in the States.

Where it really started to solidify was when my own company got blocklisted on Google Search Ads - as someone in AdTech said, it's a Rite-of-Passage of sorts. We urgently needed to find another way of generating leads and so we decided to try content marketing. I took on the task of owning and building out this function.

Since I'm not a literary genius but rather an analytical person, I wanted to create a system that would make it easy for us to create content that's consistently good enough and easy to produce. A standardized tone of voice. A blueprint for our content marketing engine.

I googled for available best practices and struck gold. My canon consisted of Mailchimp's Content Style Guide, the Associated Press Stylebook, and Scott Adams' essay "The Day You Became a Better Writer."

Since then, I tried to stick to those rules and develop a habit of writing clearly. It sometimes reads like sassy technical documentation for a slightly broader audience. But it worked for me. People complimented me on my writing. And so, I leaned in.

This was back in 2015. 

If you read one of my earliest pieces "Learning 80/20 of CustDev in 48 Hours," - also scoring perfectly on the "AI Slop scale" - then you'll notice that my writing style has remained somewhat consistent across the years:

• (2015) The State of Fine Dining (Part 1)
• (2017) A Guide to Better Writing
• (2019) Building a Remote Company: Lessons Learned from Idea to Exit
• (2021) "Mom, I'm on Spotify": How I Produced My First Music Album
• (2023) Mildly/Wildly Interesting Things About Japan 
• (2025) My Antarctic Dispatch

Lots of bullet points, copious amounts of mixed media, and Oxford commas for days.

Is AI's Writing Modeled After Me?

My writing is definitely not the pinnacle of human literature. But it is publicly available, meaning that an LLM model could easily ingest all my online writing as part of its training data.

I have published hundreds of thousands of words across multiple publications. Maybe even crossed the million word mark by now. But there are many others who have significantly more content online. Just think of Seth Godin - the GOAT - who has been writing a daily blog post since the early 2000s.

Most of us stand on the shoulder of giants. I definitely benefitted tremendously from other people's writing & guidelines when learning how to write in English. AI is no different. It learned from the entire catalogue of published words on the internet. Same principle, bigger scale. Maybe LLM's writing and mine aren't all too different. 

Just don't call me AI Slop, please. That's insulting.

The Bigger Issue That Rubs Me the Wrong Way

At this point, I want to make it clear that I'm cool with the creator of tropes.fyi

Anyone who puts themselves out there by writing online and shipping products deserves my respect. Big up, you're a real one! Keep it up.

The issue I have is the harm that AI tools can do to our perception of the world.

In fancy terms, this is called an Epistemic Collapse - in plain English: We lose our sense for (shared/actual) reality.

Maybe you even know what I'm talking about. Did you recently have a situation in your professional or personal life where you took some of the LLM outputs at face value and they turned out to be completely false?

Congrats, you ended up in Plato's Cave.

Digital technology already introduced a layer between us and reality.

AI/LLMs are doing that on steroids. Welcome to [ENTER YOUR NAME]'s Cave for every single one of us.

Scary things can happen when we stop seeing these tools for the imperfect tools that they are and start accepting them as The Truth.

• Accept incorrect metrics displayed in a B2B SaaS tool and people start losing time and money
• Accept incorrect readings in critical systems (power infrastructure, air traffic control, medical systems) and people get hurt
• Accept incorrect information over a long-enough time horizon and you might end up living in an authoritarian regime

So, watch out and have a nice Monday ✌️
Art

random thought: If you say Epistemic Collapse ​often enough, you start sounding like someone with a PhD that would make it difficult to find a job. (Good luck, AI, coming up with this sentence.)

about me:

• These days, most of my writing (about climate, energy, and security) is published on my Substack
• I wrote a book. tl;dr: How documentation allows your team to get work done in single-player mode.
• Say hi. I'm here to make friends.