Rather than giving an ambiguous score (or grade) regarding if something is written by AI/ChatGPT/Claude etc., Un•AI•ify works like the popular HemingwayApp:
1. paste in text
2. tool looks for overused patterns in AI-writing
3. gives a "rhetorical score"
Patterns include negation/affirmation ("It's not X. It's Y"), using words like only, just, really, truly, fully, etc., buzzwords, cliches, triads, and ...
The ubiquitous em dashes.
If you want a fun test, grab the blog post by the Jasper CMO, and drop it in:
https://www.jasper.ai/blog/context-ultimate-differentiator-a...
here's the analysis: