Em-dash overuse drew the most complaints in Carter Johnson’s analysis of Reddit discussions, followed by complaints about uniform rhythm and repeated sentence patterns. I found the project useful because it starts with what bothers readers. An em dash alone still tells you little about who wrote a sentence or whether it needs editing.
The project collected 89,239 posts from 47 subreddits between 2021 and 2026. Its topic filter retained 7,984, and its sampling script selected 600 longer, higher-engagement posts for review.
The published workflow assigns the review to LLM agents, although the README describes it as hand-audited. The saved output reports 604 posts, four more than the script requests. That discrepancy and the selection method limit what the percentages can tell us about readers generally.
Em-dash complaints appear in 43 cited posts, or 7.1% of the reported sample. The topic filter explicitly includes discussions of em dashes, which may help explain their lead. The final ranking also adjusts the raw counts using reviewer judgments, so its order does not follow the percentages exactly.
Read the passage around the flag
The project calls several categories “regex-blind.” In its table-building code, that means the keyword comparison has no matching rule. It does not mean software cannot measure any part of the problem.
Software can count sentence lengths and find “great question.” Deciding whether the rhythm is dull or the praise is earned requires reading the surrounding passage. A familiar sentence shape may be doing useful work.
Zero Slop’s local scorer checks phrasing alongside sentence rhythm and formatting. The AI assistant running the skill then reads the argument in context and proposes edits. Review those changes yourself; a flag is a place to look, not an instruction to delete.
Johnson’s scanner skips quoted examples for most prose rules. Zero Slop excludes blockquotes and short quoted spans from its phrase and vocabulary checks, so quoting a cliché does not count as using it. Those spans still count toward rhythm, readability, and formatting checks.
Turn each tell into a question
When sentence shapes repeat, check whether the ideas repeat too. For automatic praise, look for what earned it. Delete a polished paragraph if removing it loses no fact, reason, or useful connection.
Keep punctuation that helps and the technical terms your audience needs. An article can obey a list of style rules and still say nothing; the removal test catches what a word search misses.
Check a draft, then inspect every suggested change. Leave the original wording when removing it would cost clarity or information.