For Anyone Panicking About the New AI Detection System on Substack
I'm super interested in your comments and feedback on this one!
Today’s post is short and sweet, and hopefully sheds some light on all the controversy around AI scanning and detection here on Substack.
In case you were blissfully unaware, Substack recently announced that we can now scan our own and other people’s work to determine whether it is AI-generated.
This is the article, in case you missed it:
However, there seems to be plenty of uproar specifically related to the detection tool they call Pangram. Seems it’s not as accurate (or even remotely accurate) as users might think, as pointed out in one Reddit thread I found.
Somebody experimented with the tool on a post published in 2008, before AI even existed the way it does now, and Pangram determined that the piece was AI-generated.
Below is a screenshot of the intro post for the Reddit thread; this link takes you to the entire thread.
Where to find the tool in your dashboard:
If you’d like to test your own content before publishing a post, you’ll find the tool in your post settings, just below the scheduling tool.
I tested the post you’re reading right now and below is the result, which is 100% accurate.
What does this mean for YOU?
It means whatever you let it mean. If a tool detects AI content in something written in 2008, you be the judge on how much you trust the tool.
In the meantime, I will disclose that YES, I 100% use AI assistance for many things related to running my business and idea generation. I also use it for YouTube editing and description writing.
I do NOT write full posts using AI, but I do allow it to summarize, organize, and help me create materials for some of my offers.
As a moody midlifer who seizes every opportunity to SAVE TIME, why wouldn’t I use it?? It just gives me more time for the important things, such as refilling my bird feeders and making tea. 😆
The bottom line is this:
If you run an online business and do not use AI in this day and age, you’re missing out on having an admin assistant for a mere $20/month.
And whether you “think” you’re not using AI, even if you dabble in Grammarly, sorry, but that qualifies as AI. Probably even your Canva use qualifies.
Once again, here is the link to Substack’s post on their AI detection tool and I highly encourage you to read it if AI is a topic of interest to you pertaining to Substack writing.
Let’s have a discussion! Were you aware of this new tool? If not, and now that you are, does it concern you that someone could potentially think that your content is AI-generated even if it’s not?
"why wouldn’t I use it [AI]?" Because it can't be trusted. Because you're using tools built on the backs of other writers (like me) whose work was stolen (decades of original research). Because it uses inordinate amounts of energy and water. Because the tech oligarchs who run these companies are giving boatloads of money to support governmental efforts to defund education and libraries and dismantle the research infrastructure in the US. And they collaborate on surveillance. Because Amnesty International reports that AI companies are breaking international human rights laws.
Please be transparent. If you post AI-generated content just say so and let readers decide whether to read or not. Passing off writing or art based on others' uncredited work is plagiarism plain and simple.
If my name is on writing I wrote it for better or worse. With respect for my fellow writers and artists I don't and won't use LLMs that ingested their work.
See: Unlawful by design: Exposing the human rights costs of generative AI
https://www.amnesty.org/en/documents/pol40/0996/2026/en/
I was a writer long before computers became a staple in many homes. I wrote my first novel, entirely on a portable typewriter. If any of you are young enough to wonder where the term "cut and paste" comes from, that's exactly what writers did in early drafts, move lines and paragraphs by cutting them out with sissors and pasting them where desired. Also, no spell checks, no grammar checks and no delete or backspace option. If you typed something in error you could try to erase it, use a correction tape, White Out . . . or simply X it out and move on..
Final drafts submitted for professional work had to be perfect, which meant retyping an entire page if there was one visible mistake.
So when computers came into play, was that new technology cheating? A quip that went around at the time was "Put 10,000 monkeys in front of 10,000 keyboards and one might turn out the works of Shakespear--computers have proven this true." The inferece being that computers were allowing mediocre writers to churn out terrible writing that looked perfect. Sound familiar?
Writers now using AI are being condemned by their virtue signaling readers. I'm hear to tell you, if you couldn't discern that a long post on Substack was written mostly by AI, then you are a mediocre reader at best. I said what I said.
If a piece is written so well you are surprised by reported AI use, I can guarantee you that writer is putting far more human intelligence into the piece than Claude could ever muster.
And while on that subject, here's the real kicker. If you are using the AI dection tool, you are feeding AI the kind of polished writing that only a human can turn out, YOU are using AI AND helping train it to become undetectable.