Today’s video I go over an article I thought was AI which looks probably human, and an article I wrote with AI which an AI detector said is likely human. It’s getting impossible to tell.
© 2026 Infogalactic AG · Privacy ∙ Terms ∙ Collection notice
Substack is the home for great culture


Automated AI detectors are a fruitless endeavor unless they are kept Top Secret. Any working AI detector is immediately used to train the next AI to be more difficult to detect.
Having used Sonnet extensively for writing, I can now tell fiction that is written with Sonnet. I expect that editing will get better over time, but for the moment, I can tell that both you and Vox Day use Sonnet for writing fiction. For other LLMs, I probably can't, as I don't use them enough to pick up their "tells."