ChatGPT and AI are great time-savers, but they can easily make a lot of our writing sound the same. There are certain signs that give it away. Some of these are easy to spot, but some are a little sneakier.
We might not consciously be aware of what they are, but our brains pick up on them. Something about it just feels off, or what we’ve read sounds eerily similar to so many other things we see online.
We don’t need to stop using AI. It’s useful and helps us get more done in less time. We can, however, learn how to use it while still bringing in our own personality and our own unique insights and way of speaking.
This type of sentence structure is called antithesis. It’s used along with parallelism, which we’ll get into later. Antithesis contrasts two opposing ideas in a parallel structure. It’s a snappy way of saying something and getting someone’s attention.
Example: "That's one small step for a man, one giant leap for mankind."
I see this often in social media posts.
Example: “It’s not working harder - it’s working smarter.”
There’s nothing inherently wrong with this structure, but when ChatGPT uses it constantly, we see it so much that we equate it with AI speak. It’s a shame too, because it’s a punchy way to say things.
The first thing I like to do is to try and reword the sentence myself. That’s usually the best way to make it sound as human as possible. Sometimes I’ll flip the sentence around by taking the last part and putting it first. Using the earlier example, I might say, “Work smarter instead of just working harder.” Neither option is technically wrong, but one leans more into AI speak, and if you’re trying to avoid that, it’s worth the effort to rephrase.
If you’re short on time, or it’s a sentence that’s difficult to reword, ask ChatGPT to reword it for you and tell it not to use antitheses, contrasts and parallel structure for that sentence. You may have to tell it to do this multiple times before it understands or makes the correct change. It is something hard coded into the way it ‘thinks’ and sometimes it takes work to get it to switch to another sentence structure.
I think there are two teams in this world: Team Oxford comma and Team No Oxford comma. Really, they can both be correct. Writing is dynamic and the way we speak and write changes. This is one of those things. There are times when it’s really needed. For example:
No Oxford comma: “I’d like to thank my parents, Oprah Winfrey and God.”
With Oxford comma: “I’d like to thank my parents, Oprah Winfrey, and God.”
Makes a difference in the right context, right? Personally, I am Team No Oxford Comma unless it’s in a situation like the example above. I just think there are way too many commas floating around before ands that aren’t really necessary. It’s not about my preference though. It’s more about how much it shows up in AI writing.
Another sign is the em dash. A co-worker said this last week and it’s true: “They have stolen the em dash from us.”
I used to love the em dash. It’s such a useful tool in a sentence and makes the words feel like they’re being spoken directly to you. But now ChatGPT has overused it so much that whenever I see an em dash, all I can think is, AI wrote this.
For the comma issue, if you want to lessen the chances that your writing will sound like AI, taking out a couple of Oxford commas might help. (Gasp… clutches pearls now.)
As for em dashes, I’ve personally set up multiple GPTs with the explicit instruction not to use them, and it still does. I honestly think it’s hardwired in.
Sometimes you can just delete the em dash and the sentence works fine. Other times you can replace it with as, but, which, a comma, or even a semicolon. In some cases, you can split the sentence and replace the dash with a period.
And if all else fails, just ask it to restructure the sentence without dashes.
This is when ChatGPT groups ideas in threes for rhythm. It does this several different ways.
One way is a term called ‘tricolon’, which uses three parallel elements for rhythm and impact.
Example: “You are talking to a man who has laughed in the face of death, sneered at doom, and chuckled at catastrophe.” (from The Wizard of Oz)
It also uses parallelism, which is when two or more elements of a sentence have the same grammatical structure. This helps intensify the rhythm, draw a comparison, emphasize, or elaborate on an idea.
Example: “Their son loved playing chess, video games, and soccer.”
Again, there is nothing wrong with these types of sentence structures and they’re impactful writing and speaking styles, but ChatGPT uses them too much and it can come across as a little singsong-y.
When you see this type of sentence structure, if you feel it’s overdone in your project, ask ChatGPT to reword that sentence without using tricolon or parallel sentence structure. You may have to instruct it to do this multiple times before it makes the change for you.
Examples: discover, ensure, dive in, no fluff, unpack, dynamic, innovative, vibrant, aligns, embark, delve, keen, and many more
AI uses the most widely accepted terms and then repeats them over and over again. If that’s not the way you normally sound when writing and speaking, it can come across as a little stiff.
Use your own descriptive words. When you read over your work, ask yourself, “Would I actually say this out loud?” If the answer is no, rewrite it the way you naturally would.
ChatGPT likes to use weirdly formal transitions. While that might be great for professional reports, or legal pleadings, it’s not great for creative writing or sending an email if your goal is to sound personable and human.
Examples: moreover, furthermore, in conclusion
Of course, there is nothing wrong with these transitions, but sometimes they remind me of my high school UIL creative writing days.
Feel free to add some ands or buts in there. An of course, works as well. Using these words makes the sentence sound more natural and conversational.
Start off writing in your own words. You can even use the dictate button at the top of Word if you want to be sure it sounds like you. If you want to polish it up after you’ve written it, drop sections into ChatGPT and use this prompt, “Minimally reword this for better flow, only if necessary, but keep my voice.” If you only ask it to simply reword it, it will take over and you’ll lose some of what makes it uniquely your voice.
I like to think of writing like I think of singers. Of course, we all admire those amazing perfectly pitched singers, but we also know and love those singers that just aren’t technically perfect. The gravelly or raspy singers, or the ones who are just a little flat now and then. We love them because they have a unique sound. There is personality in their voice. It's not perfect, but it's unmistakably their own voice. That’s what writing should be.
Use AI and ChatGPT to help you write, but your writing will always be stronger when your own voice and humanity show through.
These are just a few signs your writing might sound AI-generated. If you would like more of these, let us know!