There’s an odd narrative floating around the Internet that the presence of em dashes means the content was AI-generated. There’s an even stronger counter-narrative saying this is completely false—em dashes are regularly used by human writers, too.
(Quick grammar lesson interlude: Em dashes are the longest option and look like this: —. They are different from hyphens – and en dashes –. They are used to set text apart in a sentence—like this.)
And while that debate is settled (em dashes are an integral part of writing accurately in English and, thus, used by human writers and editors as well as AI), there’s a bigger conversation at play: how can you tell if writing is AI-generated? What are the signs?
I write and edit a lot of content for the Internet. Which means I’ve used AI many times myself and have worked with other people who do the same. I’m frequently asked to edit content that was generated by AI to “sound more human,” which I’m happy to do.
As more of us rely on AI to help with brainstorming and optimizing our written content, we need to be equally aware of the common AI traps that make writing worse, not better.
AI-generated vs. AI-influenced
I think it’s important to note that there’s a difference between text that’s 100% AI-generated and text that used AI to help create the final product.
The first will be a copy-paste from whatever ChatGPT spits out, and the second involves a human writer or editor working with an AI tool to write their content. That human writer and editor can look for common AI style and phrasing and change it. A good editor will know how to catch those things and ensure the piece ends up with the right tone, message, and style for he intended audience.
Why is AI content a problem?
AI content becomes a problem when it doesn’t engage your reader. They might not be able to put a finger on it or identify that it’s AI-generated, but it’ll just sound… well, bad. If you haven’t worked to revise AI-generated content to meet its intended purpose, it’ll read as flat, stale, boring.
And those things are best-case scenario. The worst case is inaccuracy and errors. It even says right at the bottom of the screen: “ChatGPT can make mistakes. Check important info.”
The problem with AI writing is two-fold: it can be bad (i.e., not resonating with your audience) and it can be wrong.
But if you know what you’re looking for, you can fix it.
How to know if text is AI-generated: 3 signs
If we want to avoid the bad and wrong of AI-generated content, you need to look out for a few persistant cmmonalities. In my experience, here’s how to know if writing is AI-generated:
1. It’s repetitive
One reason people think the em dash is a sign of AI content is because it’s overused. (The other reason is because most people don’t use the em dash correctly, so it must be a computer. Dumb argument.)
And I don’t think it’s wrong—ChatGPT loves the em dash, but so do I.
But AI does tend to repeat itself in other ways. I’ve asked ChatGPT to write a first draft of an article at a certain word count and it will repeat the same thing multiple times to reach that limit instead of generating something new.
AI also likes to repeat specific words and phrases. You can review lists like this and this that highlight commonly-used phrases by AI such as:
- Provide a valuable insight
- A stark reminder
- Left an indelible mark
These aren’t bad phrases, but AI likes them a little too much. Keep an eye out for that repetition and familiarize yourself with those common phrases while you edit your work.
I’ve also observed some common phrasings in AI writing like:
- Lists of three (e.g., “Exercising so can make you happier, stronger, and healthier.”)
- Using the pattern, “Whether X or Y, blah blah blah.”
If you’re reading through an AI-generated piece of content, look for repetition of these writing patterns and re-write them to mix it up.
2. It’s impersonal
AI doesn’t know you. It doesn’t know your personality, writing style, brand voice, goals, values, purpose, insights, or anything that matters. Without any of this, the writing comes across impersonal and straightforward. Some text may be better that way (i.e., straightforward corporate communications), but most writing should match the tone of a real person or real company/organization (with their brand voice).
Now, you can get partway there with the right AI prompts and input to match style and tone, but you can’t account for true human insight. AI doesn’t have new ideas; it doesn’t have real experiences. But you do. And your writing should reflect that.
Impersonal writing devoid of new insights and engaging personality is a hallmark sign of AI-generated content.
3. It’s obvious
Following from #2, if you don’t have an actual person with their actual, unique insights in the writing, your AI will pull something together from what’s available on the web (i.e., what it’s “learned”). And that is going to be the most obvious, most prolific content on the Internet.
We know that there’s a ton of repurposed and/or similar content online—heck, this article isn’t even that unique! But human writers still bring their human insights and AI doesn’t. It takes what’s obvious and available and spins it into something “new.”
So if you read something and think, “uh, yeah, that’s super obvious” you may want to investigate further… it may be AI or a boring human (sorry, some human writing is also obvious).
It’s repetitive. It’s impersonal. It’s obvious. Three ways to know if writing is AI-generated... Beware.
If you don’t want to sound like a robot, come work with me—a real-life human writer and editor who knows how to capture your voice in written form. I can work from scratch or take AI-generated content and transform it into something your audience wants to read. In any care, let’s talk.
Leave a Reply