Bookshaper is in early access — subscriptions opening soon.

Guides/Feature Workflows

Proofread Mode Deep Dive

Grammar and style analysis for polished prose

What Proofread Mode Does

Proofread Mode is Bookshaper's dedicated prose-polishing engine. It zooms in on the sentence and paragraph level, analyzing your writing for grammar issues and stylistic patterns that weaken prose.

The system operates in two distinct layers. The grammar layer runs locally and delivers instant feedback on mechanical issues. The style layer uses AI-powered analysis to detect deeper craft patterns that require understanding context and intent. Together they cover the full spectrum from misplaced commas to overwritten passages.

Proofread Mode is designed for later-stage revision work. You would typically engage it after your story structure, character arcs, and scene sequencing are in good shape, and you are ready to refine the prose itself. Using it too early in the process can pull your attention toward sentence-level polish when you should still be solving structural problems.

The Grammar Layer

The grammar layer runs entirely on your local machine, which means results appear instantly as you write. There is no network delay and no usage cost. It scans for five specific categories of mechanical issues that are both common and reliably detectable.

Because the grammar layer runs locally, it works even when you are offline. You can use it on a plane, in a coffee shop without Wi-Fi, or anywhere else you write. The results are identical regardless of connectivity.

Passive voice
Identifies sentences written in passive construction where active voice would be stronger. Not every passive sentence is wrong, but overuse of passive voice saps energy from prose, especially in action sequences and dialogue beats.
Readability
Flags sentences that are excessively long, structurally convoluted, or difficult to parse on first read. Readers should not have to re-read a sentence to understand it unless the complexity is serving a deliberate artistic purpose.
Repeated words
Catches words or phrases that repeat within a short span of text. Unintentional repetition is one of the most common issues in draft prose and one of the hardest for writers to spot in their own work because the brain auto-corrects familiar patterns.
Indefinite articles
Detects incorrect usage of "a" versus "an" before words, including edge cases with silent letters, acronyms, and unusual pronunciations that trip up even experienced writers.
Sentence spacing
Identifies inconsistent spacing between sentences, such as mixing single and double spaces after periods. While modern typographic convention uses single spacing, the important thing is consistency throughout your manuscript.

The Style Layer

The style layer uses AI-powered analysis to detect patterns that require understanding prose at a deeper level than grammar rules can reach. These eight categories address craft-level concerns that developmental editors and experienced beta readers typically flag.

Show, don't tell
Identifies passages where you state emotions, atmosphere, or character traits directly instead of rendering them through action, dialogue, and sensory detail. Flagged passages are candidates for dramatization.
Cliche
Catches overused phrases and expressions that have lost their impact through repetition. This includes both common idioms and genre-specific cliches that readers of your category will have encountered hundreds of times.
Purple prose
Flags passages where the writing calls attention to itself through excessive ornamentation, overwrought metaphors, or adjective pileups. Good prose serves the story; purple prose serves the writer's desire to sound impressive.
Filter words
Detects words that insert unnecessary narrative distance between the reader and the experience, such as "she noticed," "he realized," "they felt." Removing filter words pulls the reader closer to the character's direct perception.
Adverb overuse
Identifies adverbs, particularly those modifying dialogue tags, that could be replaced by stronger verbs or eliminated entirely. Adverbs are not inherently bad, but overuse is a reliable signal of prose that could be tightened.
Dialogue tags
Flags creative dialogue tags like "she exclaimed," "he interjected," or "they breathed" that draw attention away from the dialogue itself. In most cases, "said" and "asked" with occasional action beats produce cleaner, more professional prose.
Sentence variety
Detects passages where sentences follow the same structural pattern repeatedly, such as a run of subject-verb-object sentences or a series of sentences all beginning with the same word. Monotonous rhythm puts readers to sleep.
Head-hopping
Catches moments where the narrative perspective shifts between characters' internal thoughts within a single scene without a clear break.

Working Through Issues

Proofread Mode assigns severity levels to each flagged issue, and the most efficient approach is to work through them in severity order. Address critical issues first, where the prose is genuinely confusing or incorrect. Then review warnings, where the pattern is likely weakening your prose but might be intentional. Finally scan informational flags, which highlight tendencies worth being aware of.

The filter system lets you focus on one category at a time. This is particularly powerful during dedicated polish passes. You might spend one session working through all dialogue tag issues, then another addressing show-vs-tell flags, then a third on sentence variety. Categorical focus prevents the mental fatigue of switching between unrelated types of revision.

For each flagged issue, you can accept the suggestion and revise the passage, or dismiss it if the current text is working as you intend. Dismissed issues will not resurface unless the underlying text changes. This lets you clear the noise of intentional choices and focus on genuine problems.

If you find that certain categories are not useful for your writing style or genre, you can toggle individual proofread categories on or off from App Settings. This is a project-wide preference that persists across sessions, so you can permanently silence categories that do not apply to your work without needing to dismiss individual flags each time.

Style issues are not errors. They are patterns that tend to weaken prose in most contexts, but every one of them can be the right choice in the right moment. A deliberate cliche in dialogue can reveal character. A filter word can create useful hesitation. Trust your judgment and use the flags as prompts for reconsideration, not commands for correction.