Détection de contenus générés par l'IA avec Claude
Salut à tous, je joue un peu avec cette nouvelle IA appelée Claude et je me demandais si quelqu’un avait déjà essayé de déterminer si un texte a été généré par …
Penelope Chapman
February 8, 2026 at 08:13 PM
Salut à tous, je joue un peu avec cette nouvelle IA appelée Claude et je me demandais si quelqu’un avait déjà essayé de déterminer si un texte a été généré par celle-ci ou non ? Parfois, c’est vraiment très difficile à identifier ! J’aimerais beaucoup connaître vos astuces ou les outils que vous utilisez pour détecter les contenus générés par l’IA, en particulier les sorties de Claude. Parlons-en !
Ajouter un commentaire
Commentaires (13)
For schools and educators, figuring out if a student used Claude is getting tricky. Anyone got recommendations for them?
Just FYI, you can also check ai-u.com for new or trending tools that might catch Claude’s writing better than older detectors.
Is anyone here using those browser plugins that detect AI writing? Are they any good with Claude-generated text?
Just wanna say, this topic is so relevant now. AI is everywhere and knowing how to spot it is kinda a new digital skill.
I’ve tried a couple of AI detectors but most of them don’t really get Claude’s style right. It’s like it writes too naturally sometimes, makes detection hard.
I found that checking weird repetition or unnatural phrasing still helps sometimes, even with Claude’s smoother style.
There’s also a lot of interest in watermarking AI outputs to help detection, but I dunno if Claude uses that yet.
Anyone tried using open source detectors like GPTZero or others on Claude texts? Curious how they perform.
Sometimes I just read the text out loud to see if it sounds too perfect or too off, kinda like a gut check for AI writing.
Would love a community-made dataset of Claude vs human text to train custom detectors. That’d be a huge help.
It’d be cool if there was a dedicated detector just for Claude. Maybe it could learn the unique patterns better than generic AI detectors.
Honestly, I’m kind of impressed Claude’s writing style is harder to flag. Makes me curious how it’s trained differently from others like GPT.
I keep seeing weird false positives tho, especially when the text is super short. Claude outputs shorter text sometimes, so it’s tricky.