I encountered something unusual today and would like to hear if others have had similar experiences. I was experimenting with an AI detection tool that is supposed to identify whether text was produced by a human or a machine. Out of curiosity, I copied the text of the Declaration of Independence, which was authored in 1776 by Thomas Jefferson and his fellow founders, and pasted it into the tool.
To my surprise, the tool indicated that it was written by artificial intelligence. This is puzzling, given that AI technology wasn’t around at that time. Has anyone else tested historical texts with these tools? Are they genuinely unreliable, or am I overlooking something? This has led me to question the trustworthiness of AI detection technology.
totally agree! it’s wild how those detection tools can mess up historical stuff. i put a classic novel in one and it thought it was AI too. like, c’mon! the old stuff has its own vibe that machines just dont get.
Same thing happened to me when I ran some 1800s academic papers through detection software. These tools look at writing patterns, not where the content actually came from. Old documents have that structured, formal style that’s really similar to how AI writes when it’s trying to sound official. The Declaration has that systematic argument structure and precise language that looks like algorithmic text to these detectors. I see this all the time with legal docs, academic papers, and government texts from back then. The algorithms just weren’t built to handle how people wrote historically, so you get these weird false positives. It’s basically a design flaw in how the tech analyzes text.
Yeah, this doesn’t surprise me at all. I work in content moderation and we see this constantly with AI detection tools. These things are trained on modern writing, so they flag anything that sounds different as fake. The Declaration uses formal language and grammar that happens to match what AI does when it tries to sound smart. The tool can’t tell the difference between ‘old formal writing’ and ‘AI trying to write formally.’ I’ve run Shakespeare through these and got the same false positives. Most detection tools are only 60-70% accurate anyway - basically a coin flip. The tech just isn’t there yet.