🤖 The AI Elephant in the Room: Why We Need to Talk About "Checkers"
I am going to go there! Yes, this post is all about AI.
I actually started to write this a while ago after investigating the way AI is being used. With the revelation that Hachette has cancelled Mia Ballard's book Shy Girl, first self-published in February 2025, I thought it was time to post this. I have been nervous to share, due to all the backlash that AI is getting, especially in the writing community.
In this post, I want to be open about my own usage of AI and share some important warnings for other authors. I want to investigate how you simply cannot trust anything nowadays.
🌑 The Dark Side of AI: Mass-Produced "Art"
I read some free books last year that were promoted by BookBub. One was awful; the two stars I gave it felt generous. I cannot confirm it was written by AI, but it certainly had that "vibe." Another book I read, I enjoyed—it was a good read—but after looking up the author on Goodreads, I saw they have 1,399 books credited to them.
Yes, you read that right. How does an author I have never heard of before have that many books? AI? Ghostwriters? This is where AI becomes a huge issue. People, I believe, are just mass-producing written stories, making up an author's name, and sticking it on the internet. Think of how many books you sell in a month, now multiply that by 1,399. This is the biggest problem with AI and the internet, and I do not have an answer. I guess Amazon and other ebook sites are partly to blame, but would I be a published author myself without Amazon? I don't know, probably not.
❎ A Warning: The Flaw in the "AI Checkers"
I do feel sorry a bit for Mia, and this is why. We all use AI. It's the modern world. You might not think you do, but believe me, you do. We all use modern-day docs that have spell-check on and grammar checking on—I am sorry, but that is a type of AI. Some writers use these new writing sites which are advertised to ‘Help you craft, polish and elevate your writing’. Even Grammarly, which I have used for years for spelling and grammar, is now advertised as an ‘AI writing assistance’. Modern-day writing, one way or another, uses some sort of AI.
I was on a writing platform not long ago that was running a short story competition. The organiser wrote: "I will be checking all works with an AI checker." I had never heard of this, so I went onto Google and found all these sites where you upload text and it will determine if it was written by AI. So, I tried it—only free ones, of course. I uploaded some text from one of my own manuscripts; one that isn’t finished and might never see the light of day.
The first site I used, Textguard.ai, was at the top of Google. I uploaded as much as I could (these sites have character limits). The site did its thing and, to my surprise, it said my words probably only had 37% human-written text and it was 63% sure that my words were written by AI! I was a bit perplexed. I know I wrote those words, and here was an AI detection programme telling me that they were not my words.
Shocked, I tried another site: Zerogpt.com. This time, the programme said the text was "most likely human written" but might include parts generated by AI/GPT—a 30.2% chance it was AI-generated. Okay, I thought, could the 30% be because I use spell-check and grammar check? At least this one said "most likely humans," but there was still that 30%.
🔬 The Reedsy Experiment
I didn’t stop there. The words I uploaded were copied and pasted from my Google Docs form. I then copied and pasted the exact same paragraphs from my Reedsy account. I type everything in Google Docs, then copy and paste into Reedsy to edit them using their systems.
Here are the results from that round of detection:
- Textguard: 31% Human, 69% AI. A difference, but not a lot.
- Zerogpt: Said the text was human-written with only a 14.5% chance it was AI. That is a 15.5% shift on the exact same text just because it was pasted from a different platform!
- Grammarly: Came back as 0% again.
I then decided to type it all again directly into each programme. Yes, it took a while, but I thought it was necessary to prove a point.
- Textguard: 25% Human, 75% AI written. So even though I typed this in directly, it says it is only 5-6% more likely to be AI!
- Zerogpt: Said 24.6% AI. Again, the discrepancy is there to be seen.
- Grammarly: again came back as 0%
Now, when we write on Word or Google Docs and you spell something wrong, it corrects it. It suggests commas and full stops, and we just click to change it. I love writing, but I will be the first to admit that my spelling and grammar are poor. In Reedsy, I’d once forgotten to change the spell-checker to British English. When I finally changed it, it flagged a lot of my words. You click the menu to change them all automatically—AI is involved whether you like it or not.
🕳️ The "Humaniser" Rabbit Hole
After all this, some of the sites asked me if I would like to "humanize" these words? I clicked on this and it processed something. It made me go deeper down the rabbit hole where I found other "AI humanizers." You can now create something with AI, then put it through another AI that uses words it feels a human would more than likely use. When you put this "humanized" text back through the AI checker, it says it was definitely human-written.
What a world! I tried it with the same text I had been using. It changed a few words and when I put it back through all three detection programmes, they all came back with this is more than likely human human generated. The highest 11.4% chance it was AI (Zerogpt). This means that the valid efforts of platform organisers to stop AI content are in vain. Anyone could create a short story with GPT, put it through a humaniser, and no detection programme will know the difference. What chance do we have?
🕵️♂️ The Ultimate Test: Can AI Catch Itself?
After seeing how the checkers handled my own human writing, I thought: “What would happen if I actually used AI to create a story from scratch?”
I decided to try it. I gave Gemini a specific prompt:
“Can you write a short story about a boy who dreams of being an Ice Hockey player? But he's from Australia where the weather is warm and not much hockey is played. His dream is to play in the NHL.”
The result was a 712-word story. I took this 100% AI-generated text and put it through the same three detectors. To my utter amazement, none of them said it was completely written by AI.
- Textguard: 27% Human, 73% AI.
- ZeroGPT: 76.1% AI. (High, but still not 100%!)
- Grammarly: 7% AI. (Wait... WTF?)
Think about that: Grammarly was 93% sure that a story written entirely by a computer was actually human! I decided to use a fourth site. I tried a tool called My Detector, which gave me a "Mixed" report: 11% AI, 57% Mixed, and 32% Human.
🎭 The "Humaniser" Trick
I didn’t stop there. I took that AI story and put it through a "humaniser"—a tool designed to make AI text look like a person wrote it. Once it was finished, I copied and pasted the "humanised" version back into the same detectors.
The results were even more ridiculous:
- Textguard: Suddenly jumped to a 60/40 split in favour of human.
- ZeroGPT: Now claimed the text was entirely human-written.
- Grammerly: 57% of the text appears to be AI generated.
🕵️♂️ The Final Boss: Testing the "Shy Girl" Detector
After doing all these tests, I came across an article about the ongoing saga of the book Shy Girl. It turns out that Max Spero, the founder of an AI detection firm called Pangram, was the one who provided the evidence that led Hachette to pull the book.
I thought: "Right, if this is the company the Big Five publishers trust, let’s see how I fare against them." I went through the same process as before, and the results were eye-opening:
- My Google Docs Manuscript: 100% Human Written ✅
- My Reedsy Edited Manuscript: 100% Human Written ✅
- The AI-Generated Short Story: 79% AI Written (The closest any site got to the truth!)
- The "Humanised" AI Story: 57% AI Written
This website definitely felt more accurate than the others—it didn't falsely accuse my own writing, which was a huge relief! But even this "top-tier" system was still 43% unsure about a story that I know was 100% created by a computer just moments before.
⚖️ The Verdict
This is the scariest part of the modern writing world. At any stage, any one of us could be accused of using AI to write our books. It only takes one person to put your hard-earned text through a faulty detector and get a "guilty" result like these.
What can we do? Honestly, not a lot. We have to adjust, keep our drafts and our "paper trails" (like my Google Docs and Reedsy history), and above all, be true to yourself.
🤖 So, How Do I Actually Use AI?
First of all, since last year, I have used AI. There, I said it. I use AI primarily for research, and whether you agree or not, I believe it's a valid tool. It is now 2026. In the past, there were other tools that had the same controversy:
1. The "Type-Writer" Backlash (Late 1800s)
When the typewriter first emerged, it was seen as a cold, mechanical intrusion into the intimate act of writing.
- The Criticism: Critics argued that because the machine worked faster than the human hand, authors would stop thinking deeply about their word choices.
- The Ridicule: Some editors refused to accept typed manuscripts, claiming they lacked the "personality" and "authenticity" of handwritten script (longhand). There was a genuine fear that literature would become "soulless" and "industrial."
2. The Rise of the "Grub Street" Hack (18th Century)
Long before AI, there was the "hack writer." With the expansion of the printing press, people began writing high volumes of low-quality pamphlets and "penny dreadfuls" purely for profit.
- The Controversy: Serious literary figures like Alexander Pope were horrified. They felt these "mass producers" were polluting the culture by writing for money rather than art.
- The AI Connection: Just like today’s concern about AI mass-producing books for a quick buck, 18th-century critics felt the sheer volume of "cheap" writing would drown out quality work.
3. The Word Processor and Spell-Check (1980s)
When computers moved from offices to home desks, the literary world had a collective meltdown.
- The Fear: Prominent writers argued that being able to "cut and paste" or have a computer suggest a spelling would make writers "intellectually flabby."
- The Verdict: Many claimed that the "struggle" of re-typing a whole page to fix a mistake was part of the creative process, and removing that struggle would ruin the quality of prose.
4. Even Socrates Hated... The Pen?
If you go back far enough, the most "controversial" technology was writing itself. * In Plato’s Phaedrus, Socrates argues that relying on the written word will "create forgetfulness in the learners' souls."
- He believed that if people wrote things down instead of memorizing them, they would have the appearance of wisdom without the reality of it. It’s almost identical to the modern argument that using AI for research makes a writer "pseudo-knowledgeable."
I also use AI for social media help. I’ll ask it for post ideas or strategy because I haven't a clue what to do with social media and I cannot afford a marketer. Before trying to promote my books, the only social media I had was Twitter to keep up with the Edmonton Oilers! AI helps me—not that it helps that much, as I am lucky to get 30 views on a post.
🤯 Where Do We Go From Here?
I have done so much research over the last week, and quite frankly, I do not know what to think anymore.
If even the most advanced systems in the world can be "softened" by a humanising tool, then the "AI War" in publishing isn't going to be won by software. It's going to come down to the same things it always has: Author reputation, transparency, and that undeniable "spark" that only comes from a human heart.
I’ve learned that the technology is moving faster than the detectives. As an author, all I can do is keep my "paper trail" of drafts, be honest with my readers, and keep working at my art. Because at the end of the day, a book isn't created overnight—it should be a journey.
But it does make me worry about all the genuine writers out there after these mixed results; any one of us can be accused of using AI to write our stories, all because another AI says so.
📚 What Am I Reading
Enjoyed the ride? 📖
I balance my writing life with a full-time day job, which means covers, research, and advertising all come out of my own pocket. If you’d like to see these books reach more readers—and help me stay caffeinated while I write them—you can support my work here. Your help keeps the "Stream" flowing and the books coming.
📬 Want to follow the journey?
I send occasional emails about writing, publishing, and the stories behind The Newman Adventures.
Newsletter subscribers also get an exclusive sneak peek of Book 2, An American Treasure.
👉 Join here: The Newman Chronicle
📖 Enjoyed this post?
An Irish Mystery is a fast-paced historical adventure about stolen treasures, hidden histories, and the cost of obsession.
Ebook, Kindle Unlimited, Paperback & hardcover
👉 Available on Amazon: An Irish Mystery
You can also find me on Instagram / Bluesky / X / Facbook, where I share writing updates and behind-the-scenes moments.
Comments
Post a Comment