Prose feedback

Prose feedback points out habits that are easy to miss in your own writing: words you lean on, words that come back too soon, padding, and sentences that run long. It's worked out on your device as you write. Nothing is sent anywhere, and no AI is involved.

Every check is a rule of thumb. A finding is a place worth a second look, not a mistake: plenty of good sentences use "just", the passive voice or forty words.

Open the Prose panel

In a scene, choose … (Scene details) → Prose feedback. On a phone it opens as a sheet over the scene; on a wide window it sits beside the editor, where scene details usually are, and follows whichever scene you have open.

At the top are the scene's statistics, then its most used words, then what each check found (the first 30 findings of each, with "And 12 more" after them). Select a finding to go to it: on a wide window it's selected in the text, ready to retype; on a phone it's highlighted without bringing up the keyboard.

Underline findings in the text

Turn on Underline in the text in the Prose panel to mark findings right in the scene. Each check has its own line, so you can tell them apart without relying on colour:

Point at an underline to see which check it's from. Underlines are drawn over your text, never written into it: they don't change the scene, its word count or anything you export.

Underlining starts off, and focus mode has its own switch: turning it on while revising won't follow you into focus mode. To underline there too, on a wide window with a keyboard, press ⌥ ⇧ H (Alt Shift H) while in focus mode. Outside focus mode, the same shortcut turns underlining on and off without opening the panel.

The checks

Overused words

The words the scene uses most, leaving out names (words you only ever write with a capital) and everyday words like "the", "said" and "would". Forms of a word count together, so "look", "looks", "looked" and "looking" are one word.

When a scene uses a word far more than the rest of its manuscript does (at least three times its usual rate, and at least four times in the scene), the list is headed Overused words and shows how many times its usual rate that is, such as "4× usual". Until the manuscript has a couple of thousand words there's nothing to compare with, and the list shows the Most used words instead.

Echoes

The same word again within about 50 words: "She lifted the lantern. The lantern swung." Names, everyday words, "said" and "asked" are left out, and forms of a word count together. Each finding says how many words back the last one was.

Crutch words

Words that pad a sentence without adding to it: just, really, very, suddenly, actually, basically, simply, totally, literally, definitely, certainly, quite, rather, somehow, that, and the phrases began to, started to, sort of, kind of, a bit and a little.

Filter words

Words that put a character between the reader and what's happening: "She saw the gate open" rather than "The gate opened". The list covers felt, saw, heard, noticed, seemed, watched, looked, realized, wondered, decided, knew, thought, could see and could hear, and their other forms.

-ly adverbs

Adverbs ending in -ly ("walked slowly"), where a stronger verb might do the work. Words that end in -ly but aren't adverbs, like "family", "early", "lonely" and "only", are left out, and so are the crutch words above.

Passive voice

A form of "to be" followed by a past participle: "the gate was opened", "the letters were not written". This is a guess from the words alone, so it misses some passives and catches some phrases that aren't ("she was tired" is left out, but other adjectives ending in -ed may be marked).

Long sentences

Sentences of more than 35 words.

Statistics

• Words and Sentences.
• Average sentence and Longest sentence, in words.
• Sentence variety: how much sentence lengths vary (their standard deviation). A low number means most sentences are about the same length, which can make prose feel monotonous.
• Average paragraph, in words.
• Reading ease: the Flesch reading ease score, from about 0 (very difficult) to 100 (very easy). Plain English scores 60–70; most fiction scores higher.
• Grade level: the Flesch–Kincaid grade, the US school year the text suits.
• Dialogue: the share of words inside quotation marks (“curly”, "straight" or ‘single’).

Both readability scores count syllables with a rule of thumb (groups of vowels, minus a silent final "e"), which is right for most words and off by one for some. Treat them as a way to compare your own chapters, not as exact measures.

Sentences end at a full stop, question mark, exclamation mark or ellipsis followed by a capital letter, so a dialogue tag stays with its line (“Where?” she asked.) and "Mr.", "Dr." and initials don't end a sentence.

The manuscript report

For a whole book, open the manuscript, choose … (Manuscript actions) → Prose report. It shows:

• the book's statistics and how many findings each check has across it,
• the Most used words in the book, with how often each comes up per 1,000 words,
• a Chapters table with each chapter's words, average sentence length, reading ease, grade level and dialogue,
• the Top echoes: the words that most often come back within a few lines, and the scene where each does most. Select that scene to open it with the word found.

The report reads the manuscript as it is when you open it.

Choose the checks and your own words

Go to Settings → Writing → Prose feedback (or choose Prose feedback settings… at the bottom of the Prose panel) to:

• turn each check on or off,
• edit the Crutch words and Filter words lists: separate entries with commas; phrases of several words work too. Restore the usual list puts back the lists above.

These settings are kept on each device, like the editor's other preferences.