Am I Publishing 100% AI Generated Content?

Last night I decided to try something I hadn’t done before. In the past I have checked some of our peers’ writing for AI generated content. I will only do this when others have accused (or suggested) that a specific creator is publishing AI generated content. But, until last night I had never checked out my own writing as I know that what I am writing is 100% Human written - I guess you can call me human (or am I some sort of interplanetary being that really doesn’t have a species classification?)

I am currently working on Chapter Eight of my psychological series “Voices” - some of you are familiar with Timothy. At this point I have written 206 words of the story (approximately three paragraphs).

After writing these 200+ words I decided to put it through ChatGPT for proofreading. Since I have been writing consistently for four years on Vocal, my punctuation and grammar have improved tremendously so there really isn’t much that needs to be corrected. But, since I am a stickler for things being correct I have my works proofread. I used to have a human proofreader, Irene, but due to having had brain surgery she is no longer able to proofread for me. Irene used to provide her services for free. At this point in my life, due to quitting my job in order to care for my mom full-time, my finances are very tight and I cannot afford a human proofreader. I would prefer a human proofreader so that way we can discuss my writing in detail. Therefore I use ChatGPT for proofreading, it accomplishes the job with great accuracy.

After Chat has made the needed corrections, which is usually 5% or less of what I have written. The majority of the corrections are punctuation and sentence structure. But, it is still my words left intact. An example of the changes made would be as follows (note: this example is an oversimplification):

Original: Doris said that, “John ran home.”

Corrected: “John ran home,” Doris said.

Every so often Chat does make suggestions for better wording to improve the flow in order to make for a better reading experience. I usually do not use Chat’s suggestions but will take its direction and rewrite the sentence(s) that Chat has suggested need to be rewritten.

So, last night after proofreading my story and corrections were made, the thought came to my mind to put the story through some of the AI Detection tools and see what they say. I was expecting maybe something like 99% Human written and 1% AI Generated since I do use AI to proofread my work.

I used the most popular detectors that came up on a Google search: ZeroGPT, Quillbot, GPTZero, Grammarly, Copyleaks, Scribbr, and Undetectable AI. One of them came back with a 99% Human written result. The rest gave me a 100% Human written result. This is what I expected as my story is 100% human written (or maybe Interplanetary being written).

But, then I decided to take things a step further. I did a Google search for the best AI Detector tools. I really wanted to check out the accuracy of the best tools against my own 100% Human written text.

The Google search suggested that the best are GPT Zero, Copyleaks, and Winston AI. I have already used GPT Zero and Copyleaks. Winston AI is one I had not yet used, but I have read that college professors use this tool to check their students’ work for AI generated content and plagiarism. Winston is rated as being 99.98% accurate.

With confidence I went into Winston AI expecting to see another 99 - 100% Human generated result. Afterall my work is 100% Human generated.

I was shocked with the result. It came back saying my text was 0% Human generated. Winston AI was telling me that my 100% Human written story was 100% AI generated. This assessment is 100% incorrect.

Having very little confidence in Winston AI, I decided to submit my original story before proofreading. I figured with all its punctuation and sentence structure errors it would rate somewhere in the 90 percentile range of Human written content. Boy, was I wrong. This time Winston gave my 100% Human written content a 26% Human written rating. I could not believe that this highly regarded AI detection tool could be this far off on its ratings of my work.

Needless to say, I emailed Winston AI to inform them of their faulty AI detection ratings. I have yet to hear back from them. I probably will hear back after this article is published.

I did a bit of research to find out why my 100% Human written content would rate as 100% AI generated. The following is a blanket reason:

A 100% human-written story might be detected as AI-generated by a detection tool due to a few key reasons, primarily because AI detectors are not perfect and can sometimes misinterpret writing styles that are very structured, consistent, or lack unique quirks that are characteristic of human writing.

I was given a more detailed reasoning, at this point I will stick with the “blanket” reasoning. Given this reasoning I understand why the opening paragraphs of “Voices (Part Eight)” is rated as 100% AI generated as the writing style is very basic and structured. The opening paragraphs are (almost) generic in a sense in order to lead into a more intense and in depth storyline. Like my “Blackbird Fly” series, I am writing the “Voices” series in an 8th to 9th grade reading level. This is being done as these stories are not meant to be literary marvels but instead are meant to relate to the everyday emotions we all encounter in our current day lives. My stories are heavy in character dialog but light in describing detail. I am doing this in order to leave room for each individual reader to interpret each scene in whichever way they can best relate to it.

The dialog of my characters is rendered in an everyday speaking style. Inb our day to day conversations most of us use basic language to communicate. Very few of us speak in scientific terms or use colorful poetic language as we speak with our besties. My attempt is to keep my characters and storylines real.

I guess this basic writing style can be detected as AI generated. However, on the flipside of this investigation seven of the eight AI Detection tools I use gave my story a 99% to 100% Human written story. I will trust the seven. As for Winston AI, I will not use them again and I certainly will not be giving them my money for an upgraded account. I do not trust their AI detection tools.

Note: This article was not proofread. You are getting the rough and raw version from Rick Henry Christopher!

With investigative love, RHC ❤️