Many writers cringed when ChatGPT hit the scenes, thinking we were out of a career. I know I did.

Sadly, I may have been right! Besides having to contend with imposters of the craft, we also have to grapple with the challenges of unreliable AI detector tests.

The problem comes when employers rely too much on these unreliable tests and consider the results God's truth. I had a terrible experience recently when I was a top candidate for a full-time writing position.

I had already completed three onboarding tests and had a last unpaid 3000-word test to pass before being accepted for the position. I knew the company would run a plagiarism scan and an AI detector test, and although I was concerned about the plagiarism scan, I had no worries about being flagged as AI because I honestly don't enjoy using ChatGPT. In fact, I consider it cheating and an insult to my art.

I spent a whole week researching and writing the introduction and Chapter One of a book on minimalism. I was exhausted but thrilled with the result, optimistic I'd pass with flying colours.

I eagerly awaited my appointment letter, already in a celebratory mood. Imagine how I felt when told that my writing was AI-generated and that I had lost the position! I was devastated and justifiably angry! I decided to conduct my own experiment with ZeroGPT.

Upon pasting in the test piece, section by section, I discovered that one piece was flagged as 97.56% AI-written. I knew without a doubt that this was untrue. One flagged phrase was, "The shift to a minimalist lifestyle gained popularity with the "Flower Power" generation." I had coined this using my qualitative research skills - all Google sources mentioned that the 1960s and 1970s had ushered minimalism in. I added a bit of flair to the facts. Unbelievably, ZeroGPT flagged my phrase as AI-generated!

I took my experiment further and asked ChatGPT 3.5 to rewrite an article. I pasted the text into ZeroGPT, only to be told, "Your text is human-written!" Knowing this to be false, my alarm bells were ringing loudly!

Yet, some employers are hiring and firing solely according to these results. This is grossly unjust to the writers involved. Further research revealed colleagues who had experienced similar problems. One colleague had completed work the client refused to pay for because he deemed it AI-generated. He ended his contract with her, and they parted badly. Although she was allowed to remove the terrible review, she lost her top rating on the platform, which has harmed her career.

Unfortunately, the client has all the power in these cases, and writers cannot defend themselves against unjust accusations. How can we argue with an algorithm? I decided to investigate further.

Let's examine some of the research conducted on ZeroGPT by different organisations:
OpenAI
The company behind ZeroGPT, OpenAI, claims an accuracy rate of over 98% and aims to reduce the error rate to less than 1%. However, considering my experience and the personal stake held by OpenAI, I'm more inclined to trust research done by independent sources.

Nerdschalk
In a test conducted by Nerdschalk, ChatGPT-generated texts had a high identification accuracy rate, while human texts caused some confusion. This struggle questions the overall accuracy of ZeroGPT.

Futurism
Futurism's test revealed that ZeroGPT falsely accused almost 20% of students of academic misconduct! This is a concern because it can have dire, even life-changing consequences for students!

Atheer Mahir
Perhaps the most shocking experiment is the one performed by AI Hallucination Prompt Engineer, Author, Mathematician, Programmer, and Physicist Atheer Mahir. The Biblical text from Genesis 1:1- 31 is deemed 96.96% AI-generated.

I tried another random Biblical text, Matthew 15: 1 - 35. It was flagged as 93.98% AI-generated, which we know is impossible.

And so, in scrutinising ZeroGPT, we unveil its truths and myths. As writers, we've faced the unsettling reality of being judged by AI detector tests, where the consequences of inaccuracies can be dire. The stories shared here underscore the real-world impact of false positives and highlight the urgent need for a more nuanced approach because it's evident that ZeroGPT, like its human creators, is fallible.

Employers must recognise AI's limitations in assessing writing authenticity and explore alternative methods or strategies to ensure fair and accurate evaluations. In doing so, we can move towards a future where writers are judged on the merits of their work rather than the results of a machine.

The industry needs a recalibration, recognising the limitations of AI in assessing writing authenticity.