Metadiscourse: Commentary About Your Own Writing
Interpretive metadiscourse is the aside that labels a point instead of making one: “The key takeaway is…”, “That last part matters more than it sounds”, “In other words…”. It's the author grading their own prose mid-sentence.
Why AI writes it: Models imitate the explainer register, where signposting is rewarded. But signposts multiply: when every point is announced as key, none is.
See it flagged live
Before & after
Before
The key takeaway is that pricing drives churn.
After
Pricing drives churn.
Before
We ship on Fridays. That last part matters more than it sounds.
After
We ship on Fridays — after the traffic peak, when a rollback costs least.
How to fix it
- 1Delete the label and keep the point; facts don't need chaperones.
- 2If a point truly is the crux, put it first and give it the best evidence.
- 3One “in other words” per piece, maximum — and only when the restatement genuinely simplifies.
Related patterns
Frequently asked questions
▸Isn't some signposting good for readers?
Structural signposting (headings, numbered steps) helps. Interpretive signposting (“this is important”) substitutes assertion for evidence — that's the part SlopTrim flags.