AI Phrase Checker

How it works

This is a pattern lint, not a detector. It scans your text against word lists and phrase patterns that show up constantly in LLM output: "delve", "tapestry", "testament to", "it's important to note", "not just X, but Y", "experts say", em-dashes every second sentence. The German set covers the local equivalents ("entscheidend", "nicht nur... sondern", "zusammenfassend lässt sich sagen").

Matches get highlighted in the text and grouped into a findings list with a count per pattern. The verdict is a simple density score: patterns per 100 words. The language is detected from the text itself, so you can paste German and English drafts back to back without touching a switch.

The lists come from Wikipedia's "Signs of AI-generated writing" project, which I use in my own editing workflow. They get updated as the telltale vocabulary shifts, because it does shift.

What it can't do

Tell you whether a text was written by AI. Nobody can do that reliably, and tools that claim otherwise are selling confidence they don't have. A human who writes corporate marketing copy will light this thing up, and a carefully edited AI text will pass it clean.

What the score does tell you: whether a text reads like AI output. That's the thing that costs you credibility with readers, regardless of who actually wrote it.

When to use it

• Checking a draft before publishing, especially one an LLM helped with.
• Editing AI output for a client and wanting a quick list of the obvious tells.
• Settling the "this sounds like ChatGPT" argument with something more specific than a feeling.

Everything runs locally. Your text never leaves the browser, which matters when the draft is client work.