The last pass before publishing
Proofreading software that does not argue with your style
The final pass is about correctness, not taste: misspellings, doubled words, dialogue punctuation, quotes that never close, dashes that change shape halfway through the book.
- Correctness only
- The last pass is not the place for taste.
- Both spellings
- British and American are both accepted.
- Names are words
- Your invented vocabulary is not a typo.
Proofreading is the one editing stage where a machine genuinely beats a tired human, because the mistakes are countable and the human has read the sentence forty times and now sees what they meant. It is also the stage where a general-purpose checker does the most damage, by flagging style as error.
Typos that survive every read
Doubled words, dropped spaces after punctuation, spaces before commas, three exclamation marks where one was meant.
Quotes and brackets that never close
Counted per paragraph, including curly and straight quotes mixed in the same document.
Punctuation that changes shape
Ellipses written both ways, dashes written both ways. Trivial to fix and very visible in a printed book.
Dialogue mechanics
A period where a speech tag needs a comma, and a capitalised pronoun opening a tag. The single most common error in unedited fiction.
How it goes
- 1
Upload the finished draft
Word, EPUB, PDF or plain text.
- 2
Work down the list
Each finding names its rule and its paragraph, so you can disagree with any of it.
- 3
Export
Take the corrected manuscript back out whenever you like.
Questions
When should I proofread?
Last. Every structural or line-level change introduces new typos, so proofreading before those passes are done means doing it twice.
Will it flag my style?
It should not. Fragments, dialect, unconventional dialogue punctuation and sentences beginning with a conjunction are all left alone by design.
Does it replace a human proofreader?
For countable errors it is faster and more consistent. A human still catches the wrong-word-but-correctly-spelled mistakes that no checker sees, like "form" where you meant "from" in a sentence where both parse.
What about non-English text?
Spelling is English only. Text that does not look like English is detected and spelling is skipped rather than flagging every word.