How To Detect AI Writing
Written by: Kate Williams, Executive Contributor
Executive Contributors at Brainz Magazine are handpicked and invited to contribute because of their knowledge and valuable insight within their area of expertise.
With the rising popularity of AI writing software, particularly the emergence of ChatGPT, many people are wondering whether the content they are reading was written by a person or a bot.
It’s a good question, one that isn’t always easy to answer – unless you know what to look for.
I’ve spent the better part of the last two decades honing my writing skills, first as an academic as I earned my Ph.D. in English, Language, and Literature, and then as a content writer.
In all that time, I have gotten good at spotting non-original copy. I can spot a plagiarized paragraph in a student’s essay from a mile away. I can skim a blog post and instantly tell whether the content is unique or simply paraphrasing something from another website.
So, you’d think I’d be able to detect AI writing with the same ease. But I was surprised to find out just how tricky it can be to tell when a person legitimately writes something and when a bot has created it.
And it’s only going to get worse. AI writing software will only get more sophisticated as people use it since AI learns from inputs. The more we input into AI bots, the more they improve their output.
But I don’t believe that means we are doomed to be stuck in the dark, never knowing if a human or an AI bot wrote a piece of content. I think AI will always be a beat behind humans when it comes to innovation and creativity in writing.
After doing some extensive research and even spending time using AI software, my team and I have figured out some surefire ways to tell that a computer, not a human, wrote the content you’re reading.
6 Ways To Tell Content Was Written By AI
These are some things to look for when determining whether a person or a machine wrote an article.
1. Questionable or absent research
AI bots aren’t great at conducting legitimate research. Instead, they mimic writing found online, including citing things that don’t exist. We’ve seen examples of AI bots saying things like “studies show x,y,z,” only to do our research and determine that no such study exists.
Legitimate researchers and writers will always cite their sources. If you ever come across writing that makes claims about something a study, researcher, or other source did without giving credit to the original author, you are either dealing with content written by AI or a very lazy and morally questionable writer.
2. Factual inaccuracies
Humans can make mistakes, sure. But bots seem to be better able to get away with making mistakes and not getting caught.
For example, CNET recently announced its plan to stop using AI-generated copy after several months of publishing articles with wildly inaccurate content. So inaccurate, it makes you wonder if a human was at the helm attempting to fact-check them.
For example, as the Washington Post reports, an AI-generated article confidently stated that a $10,000 deposit getting 3 percent interest would earn $10,300 annually. I’m no math wiz, but even I know the deposit would earn $300, not $10,300. So the information is almost there – the AI may have meant that the overall account would equal $10,300 at the end of the year, but that’s not what it said.
This is particularly damaging for what Google refers to as Your Money Your Life (YMYL) articles, which include information that could dramatically change a person’s financial or health status if they were to follow it.
So, if you are going to use AI-generated content, make sure to have a human fact-checking it for accuracy. Neglecting to do so could damage more than your reputation.
3. Overly formal or technical
Humans tend to write as we talk. Bots write like the humans that program them to think they should talk. That means a lot of AI-generated content is overly formal or technical, without any of the nuances of human language.
4. Inappropriately informal at times
On the other hand, some AI-generated content is grossly informal, to the point that it is difficult to gain meaning from what the bot is trying to say. In these cases, the AI bot is trying too hard to sound human, to the point that it overuses slang and conversational language.
5. Weird flow
Good writing includes the use of transitions to move from one point to another. Often, AI bots lack the knowledge of how to create transitions, resulting in choppy paragraphs that abruptly shift from one topic to another.
6. Extreme use of transitions
As with formal and informal language, transitions can be either too absent or too present. When AI bots are trained to use transitional words, they can go overboard. What we’re left with is an article full of transitional phrases such as “so,” “on the other hand,” “in contrast,” for example,” and “as a result.” When dealing with AI-generated content, these phrases are often placed inappropriately in areas that don’t require them.
There’s no question that AI writing software is here to stay. While Google figures out whether or not to penalize AI-generated content, we all need to deal with the fact that many people will use AI tools to create fast content for profit.
Knowing how to recognize AI-generated content can help you avoid paying for something that took a “writer” just a few seconds to assemble. It can also help you avoid making the painful (and costly) mistake of publishing content riddled with inaccuracies.
Kate Williams, Executive Contributor Brainz Magazine
Kate Williams has a PhD in English Language and Literature from the University of Tulsa and has been writing web content for over a decade. She founded People First Content in 2018 as a fractional content team that helps small businesses develop and implement a content strategy. She loves working with business owners and company leaders to develop content strategies that will create a lasting impact without paying for the high overhead of an in-house team.
