WHY NOT BELIEVE AI DETECTOR TOOLS WITH ALL YOUR HEART

[CharlesEBrown]

AI tends to focus on positives or overdo negatives (if ordered to), at least from what I've seen (as an experiment, a guy on Substack made AI review a movie that didn't exist, based on an AI generated poster, in the style of the movie reviews he wrote during his five or so years as a professional reviewer. It came up with something that, aside from being absolutely positive and complimentary, even when pointing out possible problems, looked exactly like his personal style). Usually that, and weird word choices that look right but "feel" wrong are the only telltale signs I always spot.
AI detectors don't know to look for those things so seem nearly random (and saw someone run the same text through an AI detector - twice in a row, and then after running a different prompt, and got 12% the first two times - don't recall if it was 12% human or 12% AI the first time, 11 the second and 34 after running a different prompt before repeating....).
So, no I don't trust the free ones and can't afford the paid ones (and would be skeptical of them too).

 

[melchi]

Is @Ai-chan writing stories for everyone in secret?

 

[Echidna_virshtall]

Real humans rarely use em-dashes. Thats usually the main giveaway right now. Current gpt loves to use emojis too.

Realistically, its probably because the ai checker websites haven't adapted yet to the current models (gpt 4.0/4.1, claude 4, gemini 2.5). Each model sort of has its own signature style, and since the websites are still using old prompts and old styles, they aren't picking it up.

I kinda do

 

[Alski]

Is @Ai-chan writing stories for everyone in secret?

Its not much of a secret, but @Ai-chan has access to everyones account and likes to post under other peoples names.

As for the em-dash debate, ive seen it used mostly by younger writers who have grown up around GPT/grammarly.