This is a keyword matching game, not a scientific AI detector. Real AI text detection is difficult, even for sophisticated tools, and no single phrase proves AI authorship. This one scores text based on patterns real research has linked to AI writing, then gives a verdict.
Overused vocabulary (delve, tapestry, realm, landscape, etc.) PlusAI — "The Most Overused ChatGPT Words" Being Ovee (Substack) — "ChatGPT's 100 Favourite Words and Phrases" Alston Antony — "300+ ChatGPT Words to Avoid (2026)" Walter Writes AI — "Most Common ChatGPT Words to Avoid in 2026" Ruben Hassid (Substack) — "Ban." yoursaislopboresme.com — "ChatGPT Overused Words List: 150+ Words" Hesam Sheikh (Substack) — "Why does ChatGPT use 'Delve' so much?" Formulaic openers/closers & hedges CleanTextTools — "Remove ChatGPT Filler Phrases" DeGPT — "The Ultimate ChatGPT Tells List: 50+ Phrases" (this one's also the source for the curly-quote and ellipsis checks) Structural patterns (em dash, negative parallelism) MEXC News, syndicated from BitcoinEthereumNews.com — "The 5 Biggest 'Tells' That Something Was Written By AI" — this is the one citing the Washington Post's analysis of 328,744 ChatGPT messages, which is where the "It's not X, it's Y" pattern comes from Nick Potkalitsky (Substack) — "Why AI Can't Stop Using Em Dashes"
