If It Smells Like Silicon…How to Spot AI Generated Content in the Wild

Artificially Intelligent and Fundamentally Flawed

Introduction

Let’s face it, AI is everywhere. You must probably be wondering if this article has a chat gpt-esque Je ne sais quoi to it. (It does not).

Two years ago, someone close to me wrote me a birthday letter on how proud they were of me and the person I was. I kept it. I believed it. Only for months later for them to admit that it was all ChatGPT. All of it. That’s when I stopped and realized it’s everywhere. HELP ME. IT'S EVERYWHERE.

Because I have been an avid LLM user I know what an unedited AI generated response looks like, or as I like to call gpt-esque response. Please keep in mind that there are people who write like this and this is not a sign for anyone to stop their natural form of expression. So here’s my top 5 ways to spot AI writing.

1. The "Not only ___ but also __" structure

This is imperial English. I used to write like this when I went to catholic school. What feels natural to me is “She writes and she also sings”, rather than "Not only does she write my lord… she also sings, your grace".

2. Some Words are Red-Flags

Words like delving, diving, dive into, a testament to, realm. I saw a list on the internet containing gpt-esque (yes, I am making this word a thing) words but then you would have to include the entire dictionary if every word sets off an AI alarm.

3. The three word pattern : __, __ and ___

My pet fish is my best friend, colleague and my dinner. You may think, "Meh, this isn’t gpt-esque...these are mere adjectives", but when you see one in the wild you will know exactly what I mean. Take this gem, for example. This is a testament to not only my grit but also my ability to be strong, hardworking and humble. To me, this sentence isn't just telling me about how the person is but rather trying to convince me about it.

4. Why Say It When You Can Ask It Back?

LLMs have a tendency to ask unnecessary questions and sound almost like customer support because of how they are trained. While this may seem supportive and friendly, it tends to add unnecessary rhetorical sentences. Instead of cutting straight to the answer, they might frame it with mini-dialogue cues like: "The result?” “So, what does this mean?” “Let’s break it down.” “The answer to this is…” This kind of delay in the information can feel like trying to mic human conversational patterns.

5. Dash on 'em

By far the most annoying and the most gpt-esque of them all. I read a sentence with something along the lines of ‘It’s not just alpha — it’s bravo’ and I know Miss GPT worked her magic.

Keep in mind that there are folks who write and this and in no way it automatically means that it’s an LLM generated text. These points are just a dead giveaway that it might be an AI.

You are probably wondering if this blog was written by AI. But given the chaotic grammar and tragic sentence structure yeah, it’s definitely all me. Not AI. Just an unfiltered human with thumbs and a Notes app.

LLMs are going to take over the world, Watch at your own risk.