Remove every em dash in one click.
Paste your text, pick what each dash becomes, copy the result. Catches em dashes (—), en dashes (–), and the double hyphen (--) people type instead. Runs in your browser; nothing is uploaded.
The cleaned text appears here.
A swap is mechanical: it cannot tell which dash wanted a comma and which wanted a new sentence. For that, and for the other twenty-four tells, the editor rewrites the sentence around each one.
A dash lets the writer skip a decision. Models never decide.
An em dash joins two clauses without saying how they relate: cause, contrast, aside, afterthought. A person picks a dash occasionally, for effect. A language model, which writes whatever is most likely to come next, reaches for it every time a sentence has two parts, because the dash is always a likely next token. That is why a page full of dashes reads as machine-written even when nothing else is wrong with it, and why one dash on its own proves nothing at all.
Wikipedia’s editors list it as pattern eight of twenty-five in their guide to cleaning up AI writing. It is marked weak alone: it counts when it has company. Read the full pattern, or check your text for all of them.