The AI writing patterns I spot immediately as an editor

Photo by Štefan Štefančík on Unsplash

I know it’s hard not to use AI for writing, especially for companies that want to produce content at scale. In fact, 74% of web pages now contain AI-generated content.

But what happens when they publish an AI-generated draft just as it was?

Well, the internet is flooded with copies that look and sound very similar. I wonder how brands stand out if they don’t weed out AI patterns before they hit publish.

The AI writing patterns I can’t unsee

Being an editor, I couldn’t help but notice common AI writing patterns that screamed, “Hey, I was written by ChatGPT”.

So, if you’re using AI to write your articles but don’t want to sound like AI, watch out for these patterns.

The EM-DASH fiasco

A few months back, the content industry was riled up because of the em-dash controversy.

There’s nothing wrong with using em-dash in writing, but it can get annoying when you overdo it. And that’s exactly what AI tools like ChatGPT are guilty of.

If you notice em-dash ( — ) being sprinkled generously, especially multiple occurrences within a sentence, that’s a strong vote for AI-written content.

How to fix: Most of the time, you can use a comma instead of an em-dash and the sentence still reads right.

Note: ChatGPT and Gemini have stopped abusing em-dash, but Claude is still persistent with it.

The colon giveaway

Colon is subtler to spot, but when you see it, you’ll find that it’s littered throughout the entire article.

I have nothing against colons, but there are times when they look out of place. I mean, you don’t need a colon for almost every explanation.

Check out the snippet below. Within the intro, there are two colons (and an em-dash).

How to fix: Rewrite the sentence so you can omit the colon.

Contrast framing

“It’s not ‘x’, it’s ‘why”

I swear I’ll scream the next time I see such blatant AI giveaway in my LinkedIn feed or anywhere.

Somehow, AI likes to stress its point by contrasting two ideas, where, in most cases, it doesn’t have to. Often, a direct statement sounds more natural (and less patronizing).

Thankfully, AI is more subtle in its contrast framing, but it’s still spottable if overdone.

How to fix: Just state the fact directly.

Repetitive Structure

This is harder to spot for untrained eyes. But if you’re an editor like me, you’ll notice that AI, specifically Claude, tends to stick with a templated style when writing.

Notice how the above subsections started with a short sentence before expanding into an actionable instruction?

How to fix: An experienced human writer would vary the style so they don’t sound monotonous.

AI Jargon

I have no better terms to describe words that AI favors, except calling them exactly for what they are — AI Jargon.

If you’re reading an article, and you can’t help but be drawn to these words, it’s likely AI-written:

• Robust
• Utilize
• Navigate
• Plethora
• Elevate

Again, these words do appear in human-written text, but we use them sparingly.

How to fix: Unless you’re writing a technical guide or corporate-styled piece, use words that you speak in everyday conversation.

The perfect article

Sometimes, there are no obvious AI writing patterns in an article, but I could still tell it’s AI-written.

I’ve been editing before generative AI became public, and I’ve never seen a ‘perfect’ article.

Humans make mistakes: a little typo here and there, syntax mixed up, or tenses slip up. But if I found a draft that reads perfectly smooth, it’s most likely an AI-generated piece of text.

How do you decide if an article is written by AI? Share your tips below.