Most skilled copywriters with years of experience in the writing industry know how to string a sentence together. They don’t need AI tools to produce quality content for their clients because they can do it themselves.
Fortunately for those of us who make a living from writing web content, many businesses don’t want to use AI tools, either. In fact, many blog publishers say that AI-written content lacks credibility, authenticity, and accuracy. As a result, they won’t accept content they believe has been written by those pesky little robots.
Enter AI Detectors…
Herein lies the problem. Many businesses and blog publishers now use ‘AI detectors’ to check whether humans or robots have written something. It’s as easy as pasting the written content into the text box and waiting for the detector to reveal the percentage of likely AI content. The higher the percentage is, the more likely it is that AI has been involved.
But are they accurate? No. At least not in my personal experience. I can’t count the number of times publishers have asked me to resubmit new content because my pieces have been AI-generated. Has my client content ever been AI-generated? No.
You see, it all comes down to the formal nature of the content. The more fact-based it is, the more ‘AI-like’ it can sound. Apparently.
Examples of AI Detectors Getting it Wrong
The Treaty of Waitangi is our founding document prepared in 1840. After pasting the first three articles of the Treaty of Waitangi into an AI detector, we’ve now just learned that AI did, in fact, exist in 1840. The yellow indicates the pieces of the content that are AI/GPT generated:
Artificial intelligence apparently also existed in the 1770s when the Declaration of Independence was written and signed:
AI was also around when various biblical people wrote the Bible:
How Do AI Detectors Work?
AI detectors aim to identify whether a human or an artificial intelligence system wrote an article, blog post, or something else. The text is analysed using natural language processing techniques and machine learning algorithms to assign a ‘score’ on the likeliness of AI writing the text.
Some AI detector businesses describe themselves as ‘highly accurate’, but that may not actually be true. I just penned this entire article from my own brain in a Word document, and apparently, I used AI for nearly 17% of it.
Comments