Back in September of 2025, I wrote an article titled "Does Your Writing Look AI-Generated?". It seems like ancient history now, given all the changes in AI since then. Back when I wrote that article, I tried to narrow it down to 5 signs and fixes, but there are so many now that I think that’s an impossible feat. However, I took my best stab at narrowing it down to an additional 10!
Here’s a quick list of those signs from the first article:
Most of those ‘tells’ are common knowledge now, so I felt it was past time to write a part 2.
Let’s get into the signs and what you can do about them. Also, just because someone uses any of these in their writing does not necessarily mean they've used AI. AI (LLMs) were trained on a variety of professional writing, and these writing styles have been used for many years. The problem is overuse and AI's inability to tell when it's gone overboard.
These change from time to time. I wrote about this in my first article when AI constantly used the words ensure, explore, and discover. Now it’s hung up on the word quietly or quiet. It shows up in almost everything I see now, and half the time, the addition doesn’t even make sense.
It shows up in articles, social media posts, speeches, and even in YouTube titles and scripts. I can’t unsee it now.
Here is a list of words and phrases I see constantly right now:
There is nothing wrong with these, but when they’re used too often and in everyone's writing, they become an AI ‘tell’.
Everybody wants to have their own mic drop moment at some point in their lives, but AI really hogs the stage.
Example:
“The professionals who act now won’t just keep up. They’ll lead what comes next.”
Not every sentence needs to be epic and using too many of these gets exhausting. I’ve also noticed lately that AI likes to write one-sentence paragraphs, and it’s usually when it’s trying to add in a mic-drop moment.
AI loves to be super dramatic, and one way it does that is by splicing sentences and making them sharp and punchy.
Example:
“The deadline is approaching. The team is stressed. Mistakes are happening. Something needs to change.”
I saw this gem of a comment on a YouTube video I watched recently:
AI also likes to spit out one-word sentences.
Example:
“Clear. Concise. Controlled.”
As an aside, remember when people used to respond like this on social media?
"SAY. IT. LOUDER. FOR. THE. PEOPLE. IN. THE. BACK."
It was even worse when they added the clapping hands emoji between each word. Those one-word sentences give me the same type of feeling.
To be fair, it’s fine to have some short and punchy sentences. They give emphasis. The problem is when they’re used so frequently that it makes the writing much more difficult to read.
ChatGPT is the worst at metaphors and analogies. It really tries hard to be clever. Sometimes what it gives you is entertaining, but other times it seems forced.
Example:
“The stale office coffee tasted like a Monday morning that had already given up.”
Sometimes it’ll even mix metaphors, which I really can’t judge too harshly, since I am a frequent offender. Maybe some of my writing ended up in the pile AI trained on? (Oh, the horror.)
ChatGPT specifically loves to bold random words. I don’t know why, but when I see a weirdly bolded collection of words for no reason, I know Chatty did it.
AI has a real aversion to using contractions. No, you don’t have to write with contractions, and professional writing probably shouldn’t use many of them, but I think the quickest way to sound like a robot is to not use any contractions ever.
Try this experiment: Right now (if no one is looking), pretend you’re in grade school and hanging with your friends on the playground, and you’re all talking like robots to each other. Just without thinking, talk out loud right now like you’re there and pay attention to what you sound like. You’ve stripped out all the contractions, right? Told you.
I don’t know the proper term for this, but I know it when I see it.
Example:
“The meeting started with a review of the numbers from last quarter. Everyone shared their thoughts about what had worked so far. The team then discussed what they wanted to change next.”
It’s so boring that I almost fell asleep.
AI has a sort of rhythm to the way it sounds. It either uses super short, choppy sentences, or every sentence it writes is roughly the same length. Humans don’t generally write or speak that way. We vary sentence length. Sometimes when reading AI writing, you don’t explicitly notice this, but you know something feels off.
Alright, now we’re getting technical. Don’t quit on me now, though. Once you see this pattern, you won’t be able to unsee it.
A conditional opener starts with a dependent clause that sets up the condition or situation for the main idea.
Example:
“When firms improve their workflows, teams can spend more time on client work.”
In the example above, “When firms improve their workflows” cannot stand alone as a complete sentence, so it leads into the independent clause, “teams can spend more time on client work.”
Again, we need these types of sentences, and there is nothing wrong with them. AI just uses them too frequently.
This one is pretty easy to spot, so I won’t give it much of a preamble, but to reiterate, there is nothing wrong with this structure other than using it in every single paragraph of a written piece.
Example:
“Firms face one major challenge: keeping up with rapid technology changes.”
or
“The message is clear: firms that embrace AI now will be better positioned to adapt, compete, and lead in the future.”
Have you heard of uncanny valley? It’s the phenomenon where a computer-generated figure or human-like robot very closely resembles a human being but arouses a sense of unease or revulsion in the person viewing it.
Reading AI gives you the uncanny valley feeling. It’s just too perfect. You know something is off, but you can’t figure out what it is.
Human writing isn't generally perfect. That's what makes it interesting. Now, I’m not suggesting you make mistakes on purpose. I’m saying that, as humans, we naturally vary our sentence length, use incorrect grammar sometimes, mix metaphors and some of us leave out Oxford commas because we don’t like them.
All this gives our writing personality, and it’s possible to show personality in our writing and still sound professional.
I think this is the most important tip, but I also think that no one wants to hear it. We’re all busy and AI ‘sometimes’ helps us work much faster. But I promise this will help you in a myriad of ways. It’ll make your writing more human, more personable, more individual and it’ll help your brain stay in shape.
Try to write using only your brain first, even if it’s just a very messy outline or first draft. A study done at MIT showed, “Brain connectivity systematically scaled down with the amount of external support: the Brain‑only group exhibited the strongest, widest‑ranging networks, Search Engine group showed intermediate engagement, and LLM assistance elicited the weakest overall coupling.”
There are even more studies on this and if you’d like tips on how to keep your mind sharp under AI's influence, you’ll want to check out this article: Don’t Let the Robot Make You Rusty
Run your writing through other AI tools. There are many, and some are better for certain things, but it’s good to have options. See what sounds more natural. Grammarly, QuillBot, and Claude are some options. I’ve been told that Claude (Fable) sounds much more human, but since I haven’t used it myself, I can’t confirm.
If you’re unsure how to start your writing project and really need to ask ChatGPT to write something for you first, make yourself retype what it gives you. While you retype it, sometimes your brain will automatically reword it in a way that sounds more natural.
Another option is to read it out loud. Is it hard to read? Would you say it that way? Do you normally use those words?
Edit your settings in ChatGPT's backend. You can improve (although not completely) its writing by adding some parameters.
Also, if AI writes something that sounds off, ask it what type of sentence structure it used. Keep a list of these types and add them, with notes, to your preferences within your AI tool. Of course, some of those sentence structures don’t need to be completely avoided, but you can ask AI to lower the frequency.
Another trick is to take some of your writings (that you’ve written completely on your own) and upload them to your AI to train it to write in your voice. It definitely won’t be perfect, but it can improve things.
This is old advice, but it’s still suggested because it works. Write what you can, then set it aside for an hour, or even a day if it’s a longer piece. Come back to it later, and you’ll be surprised by what you notice.
AI is your biggest fan. Oh, you thought your mom was your biggest fan? She has nothing on AI. ChatGPT specifically is very ingratiating, and it’s what is referred to as AI sycophancy.
Example:
“The fact that you spotted this means (something super fabulous about you)”.
or
“You’re right to question that (then adds that it completely changed its mind because you’re so smart)”.
I’ve added this last tip so you can watch for it. AI is a confirmation bias machine, so be aware of that when getting feedback from it. If you want tips on how to combat this, you’ll want to check out this article: How to Argue with AI
It’s unfortunate that we’ve gotten to the point where we need to be careful about saying or writing certain things for fear of sounding AI. For example, we often use parallelism when we speak and write (i.e., it’s not this, it’s that), but now it’s an AI indicator. At this point, many of us are extracting all of the em dashes from our work, and now my precious bullet points are being threatened. So, where is that line? At what point should we stop changing our speech and writing patterns just to avoid sounding AI? I think we have to find that middle ground.
If you think I missed any signs in this list, check out Part 1, as I may have included it there. If not, I’d love to know what your AI 'tells' are.