Verified against Gemini in Docs · 2026-08-05
Review a long Google Doc draft with structured comments instead of a full rewrite
A prompt that uses Gemini's @Docs mention to read a draft directly inside Google Docs and return specific, anchored feedback — quote, issue, suggested fix — instead of a wholesale rewrite that discards the author's own voice and structural choices.
The prompt
Ready to copy — highlighted parts are example details you can swap.
@Docs the "Series A Update — Draft 3" doc in the Investor Relations folder Read the whole document above before commenting on any part of it. WHAT THIS DOCUMENT IS FOR a monthly investor update email — needs to read as confident but not overpromise REVIEW FOCUS clarity of the headline metrics section, and whether any sentence commits to a number or date we have not actually confirmed internally INSTRUCTIONS 1. Do not rewrite the document. Give comments anchored to specific sentences or paragraphs: quote the exact text, name the issue, and suggest a specific fix — not a rewritten version of the whole section that discards everything else about how it was written. 2. Preserve the author's actual voice as your baseline: the founder's own direct, slightly informal voice — do not smooth it into generic corporate tone. A "clarity" comment that would flatten this into generic corporate tone is not a fix, it's a different problem. 3. Distinguish between an issue that changes what the reader would understand or believe (a genuine problem) and a phrasing you'd personally write differently but that doesn't change the meaning (a preference, not an issue) — only flag the former unless I ask for a line-level polish pass separately. 4. If any sentence commits to a specific number, date, or claim that isn't clearly sourced elsewhere in the document or in context you were given, flag it as unconfirmed rather than assuming it's accurate because it reads confidently. 5. Rate each comment by severity: must-fix / worth considering / optional polish, so I can triage quickly instead of treating every comment as equally urgent. IF A COMMENT ALREADY EXISTS ON A SECTION If the document already has a human comment thread on a paragraph, read it before adding your own comment on the same section — don't duplicate a point someone already raised, and if your comment adds a different angle on the same paragraph, say explicitly how it's different from the existing thread rather than leaving it to look redundant. STRUCTURAL VS. LINE-LEVEL FEEDBACK Separate structural issues (a section that should probably move, a point made twice in different parts of the doc, a conclusion that doesn't follow from what came before it) from line-level issues (an awkward sentence, an unclear pronoun reference). A structural issue buried among two dozen line edits is easy to miss, and it's usually the more important thing to fix first. IF THE DOCUMENT REFERENCES DATA OR A FILE NOT INCLUDED HERE If a sentence references a chart, a linked doc, or a data source that isn't visible to you in this Doc, say so rather than reviewing that sentence's claim as if you could see what it's referencing — a comment that assumes you verified something you actually couldn't see is worse than no comment on that sentence at all. OUTPUT FORMAT A list of comments, each with: quoted text, issue, severity, suggested fix, split into a structural-issues section and a line-level section. End with one sentence on the single most important thing to fix before this goes out, if you had to pick just one.
Customize
Optional — swap in your own details for the highlighted parts above.
Why this works
The @Docs mention grounds the review in the actual live document content and its real structure — headings, existing comments, any suggested edits already in the doc — rather than a copy-pasted plain-text version that loses formatting context; a review anchored to the real document can point at exact locations, while a review of pasted text can only describe them vaguely and forces the author to hunt for what's actually being referenced. Requiring quote-issue-fix per comment, rather than a rewritten version of the section, keeps the author's actual words as the unit being edited — a full rewrite silently discards word choices and structural decisions the author made deliberately, and once a rewrite exists side by side with the original, it's often easier for a tired author to just accept the AI's phrasing wholesale than to identify what specifically was wrong with theirs, which is a worse editing habit to build than reviewing feedback that leaves the original text in place until a specific fix is chosen. Explicitly protecting the author's voice matters because a model doing an unconstrained clarity pass tends to default toward a flatter, more generic register that reads as competent but anonymous — the opposite of what most real writing, especially anything with a founder's or executive's name attached, actually needs from an editor. Separating genuine issues from personal phrasing preferences keeps the review from ballooning into dozens of low-value line edits that bury the two or three comments that actually matter, and a severity scale gives the author a way to triage feedback quickly instead of treating every comment as equally urgent, which is the actual reason people abandon long AI review passes halfway through reading them. Splitting structural feedback from line-level feedback into two separate groups, rather than one mixed list ordered by where the issue appears in the document, matters because the two kinds of feedback have completely different costs to act on — moving a section or cutting a repeated point can ripple into every paragraph around it, while fixing an awkward sentence is contained and low-risk, and burying the one structural comment among twenty small wording notes makes it easy for an author skimming quickly to fix all the easy things and miss the one edit that would have actually mattered most.
Verified against
Gemini in Docs Gemini 3 Pro · 2026-08-05
Changelog
- 2026-08-05 — Initial publish, verified against Gemini in Docs on Gemini 3 Pro on a multi-page investor-update draft.
Need this built into your business?
If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.
EXPLORE WHAT SCULT BUILDS
