AI Detection

The Navajo always wove an error into each quilt or rug they made. They reasoned that if God was perfect, people being perfect was an insult to God as it was pretending to be like God. An error would therefore be the mark of humanity.

In the same way, AI detection software is finding that if an indie author has good proofreading, or access to it, that author's grammatically perfect production is "AI-generated" which could be said as "70% AI-generated" or "70% likely to be AI-generated" with the number before the percentage varying wildly depending on time of day, or which program, or whatever; it's somewhat random according to most accounts. While the vast majority of observers consider such numbers and this software in general to be a total fraud, meaningless and utterly useless, some writers are actually concerned that if software identifies your writing as potentially AI generated, that can be bad for you, so you should do whatever you have to to prevent that or redirect the software. But what to do? Make bad grammar? Unusual grammar? Make up your own words? Good question.

One obvious solution comes from the Navajo. Just slip a little imperfection in there. If a typo makes it clearly not AI-generated, make a typo. Better yet, put the same typo in every paragraph, or even every sentence. Make your own distinctive typo (for example, always spell "the" as "th") so that the reader is regularly if not constantly reminded that a real person generated this sentence.

Now I find this soluution fascinating, yet I haven't done it yet. That's partly because I haven't published anything since I became aware of this problem. When I do, I always tell Amazon that AI did not generate any of my work (although I am a little unsure of the definition of AI when it comes to the pop art I use on my covers). But telling Amazon, I figure, is meaningless. So is putting a clear statement in the front of the book that says A real person wrote this book. AI could generate this statement; AI could publish everything I do. Why would it be worried about telling the truth? It simply does as you tell it.

Back to my solution: every author inserts a flaw into his/her work, ddeliberately, as a kind of signature. The "AI detection" which trips up on anything, as long as it's good grammar and proofread, now clearly recognizes your work as. "not AI." Not even a sentence of it? Do you have to put a flaw into every sentence?

Here's a world where typos are not errors, but rather an urgent attempt by real people to coommunicate to other real people. It's a wild world, isn't it?