01/06/2026
We're all using AI to help us in our work, daily lives and planning. I know there's still lots of new people that are going to be using LLMs for the first time. What you might not realize is that there's several dead giveaways that your text was written by a robot and not a human. Here's a couple tips to avoid AI language model tropes, so your text can look a little less AI and a little more human.
1. Eliminate Em-Dashes. Most people (outside of professional writers) don't even know what an Em-dash is (it's the extra long dash: —) or how to type one on their keyboard. When you see text that has multiple Em-dashes per paragraph, it's basically a dead giveaway that the text was written by a robot. And it's too bad, because the em-dash is actually a great part of English punctuation often used by professional writers, but it's overuse in LLM written text has now made it one of the biggest signs of AI text.
To avoid using em-dashes in your copy, either replace them with different punctuation like commas, or you can tell your LLM to not use them at all with code like this:
root ::= content* content ::= allowedchar allowedchar ::= [\u0000-\u2013\u2015-\uFFFF]
2. Avoid overly elaborate, fancy or cliched terms. When you see the following words used in text copy, it's easy to assume that the copy was written by a LLM:
- intentional
- gentle
- underscore
- unwavering
- elevate
- amplify
- pivotal
- performative
When you use an LLM for writing copy that's displayed publicly, don't use words that you wouldn't use in everyday speech. If your written copy sounds like it was written by a poet who consumed too many mocha lattes and went crazy with a thesaurus, no one is going to believe that you wrote that text yourself. You can avoid this by telling your LLM to not use overly fancy or cliched words, and to speak in plain, informal language.
3. Sentence structures that set up and then negate or expand beyond expectations (like 'X isn't just about Y' or 'X is more than just Y'). This is another over-used LLM technique that immediately gives away that your copy was written by a robot and not a human. Once you see that it's a regular trope of AI text, now you'll probably start seeing it everywhere.
You can avoid this by telling your LLM: "I want to avoid any sentence structures that set up and then negate or expand beyond expectations (like 'X isn't just about Y' or 'X is more than just Y'). Instead, I want you to use direct, affirmative statements. I also want you to be creative with your sentence structures and expression styles."
4. Overuse of Bullets and Numbered Lists. Bullets and numbered lists are a great way to organize blocks of text into easier to understand points. But when you see multi-tier bullet lists all throughout text, you can pretty much assume it was AI generated. Using bullet and number lists is fine but generally it's better to write in full paragraphs, only breaking content into lists when necessary. I would also recommend using a single tier bullet list at this point, as the multi-tier lists are more of an AI giveaway.
Everyone is going to use AI to help us in our daily lives and businesses. But the content you post publicly should reflect your personality and actual, real-world presence as much as possible. Hopefully this 100% human written text can help your LLM generated copy to sound more grounded and be a true reflection of your personality.
