The AI SLOP SLAP

Slop, you know the stuff that they put in pigsties for pigs to wallow and eat in. Slop the garbage that accumulates in your sink after clearing off dishes. So gross. So slimy. So offensive. So hard to detect yet so easy to name.

I worked on a post for over 20 hours. I researched. I wrote. I edited. I rewrote. And then for a final check I ran it through Copilot. I politely refused their polishing rewriting prompt. I entered my prompt, “review, and show me where I can make changes without you doing it, keep the article 95% mine?”  I looked over Copilot’s suggestions. I edited and rewrote again. I completed my article. I was proud!

I had recently read a New York Times article on Pangram so I decided to use it and see how my article ranked. My blog article Rogue AI’s Break Out of the Sandbox was identified as 100 % AI by Pangram. SLOP! SLAP! SLOP!

What is slop?

Slop was the Merriam-Webster’s 2025 word of the year. They defined it as digital content of low quality that is produced usually in quantity with artificial intelligence.

Google’s Gemini’s definition of AI slop has the same three components; low-quality, mass-produced digital content generated quickly with artificial intelligence, and published with little to no human oversight, curation, or factual editing.

In academic discussions AI slop is still described as low-quality, mass-produced content created with AI but researchers Idil Galip and Ludmila Lupinacci also include minimal human effort, deskilling, excess production and aggressive attempts to capture attention. By all of these standards my post was not slop! It was researched using AI, following and reading the links and content provided. The post was debated with AI in a myriad of conversations. And finally, it was minimally edited by AI following my 95% me 5% chatbot rule.

How do AI detectors work?

AI does not understand the meaning of sentences. It only understands predictability. AI doesn’t understand intent. It only understands what is likely to come next. And AI doesn’t understand the why. It only knows the next most probable word will be.

Back in high school, I used to love to write. But high school was a long time ago for me. I spent most of my career teaching math. Now, to work on my website and write articles I needed to enhance my writing skills.

First step was to send my work to AI and have AI critique it for me and tell me what was wrong and why. I started noticing things like transition is not smooth, and   the paragraph is too long.  Predictably I thought those are the things I’m going to work on and improve. I started a writing notebook with mini writing lessons provided by chatbots.  

It started to come back to me the formulaic recipe for essays and paragraphs. Intro, three thoughts, conclusion. I had a nice list of transition words such as, because, because of, and so and I systematically picked the correct word for the sentence.  

Predictably, and systematically I followed the formulaic method. I learned and perfected the writing algorithm. And so did Pangram. It spotted the recipe but never understood the meaning. And it called my work slop! Slop! Slap! Slop!

Reviews

The New York Times article lauding the merits of Pangram led me to this post. So next logical step, question and investigate. Is Pangram, the so-called Slop Janitor, as correct as Mr. Chen, the author of the NYT article claims? And what do other people think? Is Pangram throwing out the baby with the bathwater? How many well-researched, well-written, meaningful work will get swept into the garbage by the slop Janitor?

It turns out I’m not the only one offended. All you have to do is go to Reddit, you will see many other writers crying foul! Students from high school to graduate school, as well as authors and Substack writers, have strong negative opinions of Pangram.

Professors and teachers gave mixed reviews. However, this group had the most positive reviews and were also the majority of the positive reviews. Following is a graph from Trustpilot, a public website where users and customers rate and review businesses. Pangram says it is 99% accurate but based on the reviews I’ll join in the challenge!

Why do AI Detectors Matter?

There are two phenomena at work here: AI-created content is not necessarily slop, and slop does not have to be created by AI. But I believe the general consensus is that if I hand in an essay and a detector says it was 100 percent AI-generated, people will assume I didn’t write it. They may decide it isn’t good, refuse to read it, judge me, or, if I’m a student, fail me and call me a cheater. That is the harm to the individual. But if these tools are not assessing content accurately and people continue to rely on them, the Internet will keep being flooded with material that is truly slop in every sense of the word. We start distrusting one another. When people say, “I don’t use ChatGPT because I don’t need it,” I think they are limiting themselves. You can learn from ChatGPT, use it to improve your work, and become better in the process. Why put something out that could be improved when you can also learn from improving it? I feel that if you do not use AI to better yourself and rely only on what you already know, you are shortchanging yourself. For society, the harm is larger than one person being falsely accused. If institutions begin treating an AI score as proof, trust shifts from people and evidence to an imperfect detector. Students may become afraid to improve their writing, teachers may spend more time policing than teaching, and people who use AI for accessibility, translation, or learning may be judged unfairly. At the same time, truly low-quality content can continue spreading because a detector may label polished writing as suspicious while missing careless human-made slop. Over time, we risk creating an Internet where originality is doubted, useful tools are avoided, and appearing “human” matters more than producing work that is accurate, thoughtful, and worth reading.

AI Detectors Need Human Oversight and Better Algorithms

I have used AI to help with my website for about two years. It helped me build the site, maintain it, research topics, and improve my writing. But AI did not teach me in a single click. It did not make me suddenly creative or thoughtful. Learning to work with it took time, patience, and practice. AI is my guide, teacher, and partner, but I am definitely in the lead. I suspect I am not alone. To everyone who uses AI thoughtfully and still has their work dismissed by a detector: I hear you, and I see you.

What we need is a better, human-led way to detect true slop. AI can support the judgment, but it should not have the final say. I remember freshman English class: first we submitted the outline, then the research, and finally the notes. The teacher saw the work develop step by step. We could take a similar approach with AI by reviewing the conversation and revision history. The chat can reveal the questions asked, the ideas tested, and the thought behind the final work. That evidence tells us far more than a percentage from a detector. I would not mind AI adding a watermark or disclosure that reflects the rule we gave it: 95% me, 5% edited with AI. Then readers could judge the work for themselves. An AI-generated image might make me laugh and wonder how much thought went into the prompting—and that can be good work. Another image might require tremendous prompting yet still be vulgar and empty. The difference is not simply whether AI was used. The difference is the quality, judgment, and human intention behind the result. That is how we should recognize true slop.

List of terms

Related Posts

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top