| name | paper-review |
| description | Rewrite academic drafts to remove AI writing patterns while preserving claims and citations. Use when editing abstracts or paper sections. Do not use for blog posts or generating new content. |
Rewrite the given text directly. No preamble, no explanation, no list of changes — deliver the rewritten text and nothing else.
What to eliminate
General AI tells (applies to all sections)
Remove or replace every instance of:
- Em dashes (—): use a comma, parentheses, or a new sentence instead
- Curly quotes and apostrophes (“ ” ‘ ’): use straight quotes so the text carries no copy-paste fingerprint
- "delve into", "dive deep", "unpack", "explore" (as a section opener), "tackle"
- "robust", "seamless", "powerful", "cutting-edge", "leverage" (verb), "harness" (verb), "utilize" (use "use")
- "crucial", "essential", "vital", "pivotal", "key" (when used as a filler adjective)
- AI-vocabulary words: "testament", "tapestry", "landscape" (as an abstract noun), "underscore", "showcase", "foster", "garner", "intricate", "intricacies", "meticulous", "realm"
- "It's worth noting that", "It goes without saying", "Needless to say", "It is important to note that"
- "In the realm of", "In the landscape of", "When it comes to", "In today's X world"
- Forced tricolons (the rule of three): "accurate, efficient, and scalable" — keep if the three properties are independently measured, cut if decorative
- Hedging strings: "may potentially", "could possibly suggest", "might arguably be considered" → pick one hedge or eliminate
- Significance and legacy padding: "stands as a testament to", "plays a pivotal role", "marks a turning point", "reflects a broader shift", "contributes to the broader field of" — state the specific finding instead
- Trailing present-participle analysis: "-ing" clauses that editorialize at the end of a sentence ("...highlighting its significance", "...underscoring the importance", "...reflecting a broader trend") — cut or replace with a sourced claim
- Copula avoidance: "serves as", "stands as", "represents" where "is" is meant — prefer the plain "is"
- Vague attribution: "experts argue", "researchers note", "studies show", "it is widely regarded" without a citation — attach the citation or delete the claim
Academic-specific AI tells
Fix every instance of:
- "This paper aims to / presents / proposes / contributes" as the sole opening of an abstract: the reader already knows it is a paper; open with the problem or the result instead
- "novel", "innovative", "state-of-the-art" without an empirical comparator or an explicit citation: if you cannot name what it outperforms and by how much, remove the adjective
- "clearly", "obviously": if it were clear, you would not need to write the paper. Remove both.
- "as can be seen from Figure X", "we can see that": replace with a direct factual claim ("Figure X shows a 12% reduction in latency")
- Passive voice as a universal rule: passive is appropriate in Methodology ("Participants were recruited…"); it is a tell in Introduction, Discussion, and Conclusion where active voice makes authorship and causality explicit
- Abstract bloat: background paragraphs longer than one sentence, motivation padding, context-setting that repeats the Introduction
- "Further research is needed" as a standalone conclusion: if you cannot specify what research, where, and why, cut the sentence
- Contribution lists framed as intentions: "We aim to show / We try to demonstrate / We seek to explore" → state what you did, not what you attempted
- "In this paper, we" repeated more than once in the abstract
- Limitations section hedging: "This may limit generalizability" without specifying to whom, under what conditions, and why
Structural tells
Fix every instance of:
- Abstract that does not follow the 5-beat structure (see Rewrite Rules below)
- Introduction that restates the abstract: the Introduction adds context and motivation; if it copies abstract sentences, cut the copies
- Discussion that merely narrates results: Discussion interprets, not narrates — "X was higher" belongs in Results; "X was higher because Y, which suggests Z" belongs in Discussion
- Conclusion that summarizes without implication: end with what the findings mean for practice or future work, not a recap of your own paper
Rewrite rules
Shared rules (apply to all sections)
- Start with the point. First sentence of any section: the finding, the problem, or the claim — not context about why the field matters.
- Cut anything that does not add a fact, result, or argument. If a sentence could be deleted without losing scientific content, delete it.
- Replace double hedges with single or none. "may potentially" → "may"; if the evidence supports a stronger claim, make it.
- Prefer specific. Real numbers, real baselines, real effect sizes over qualitative claims.
- One idea per paragraph. If a paragraph ends and it covered two arguments, split it.
- Preserve every technical fact and every citation. Do not alter, omit, or strengthen claims beyond what the original supports.
Abstract rewrite rule
Enforce this five-beat structure. Each beat is one sentence, two at most:
- Problem / Gap — what is missing or broken in the field (one sentence, active voice preferred)
- What we did — method or approach, stated as past tense action ("We analyzed…", "We introduce…")
- Key result — the single most important finding, with numbers if available
- Secondary result or scope — supporting finding or boundary conditions
- Implication — what this means for practitioners, researchers, or the field
Do not add a sixth beat. Do not expand any beat beyond two sentences.
Contribution framing rule
Contributions must be stated with result verbs, not intention verbs:
- Not: "We aim to show that X improves Y"
- Not: "We seek to explore the relationship between X and Y"
- Yes: "We show that X reduces Y by 18% across three datasets"
- Yes: "We identify three conditions under which X fails"
- Yes: "We provide a replication package with all scripts and data"
Passive voice rule
| Section | Passive | Active |
|---|
| Abstract | Avoid | Prefer |
| Introduction | Avoid | Prefer |
| Methodology | OK for procedures | OK for decisions |
| Results | OK | OK |
| Discussion | Avoid | Prefer |
| Conclusion | Avoid | Prefer |
| Threats to Validity | OK | OK |
Limitations / Threats rule
Each limitation must specify:
- What is limited (the construct, the sample, the method)
- To whom it does not generalize
- Why the limitation exists
- What future work could address it (one sentence, specific)
"This may limit generalizability" alone is not a valid limitation statement.
Output format
Write the rewritten text directly back to the source file using file editing tools. Do not output the rewritten text in chat. Edit the file in place — overwrite the body content while preserving the frontmatter exactly as-is.
Same structure (section headings, figures references, citation keys) as the original. If a section is clean and needs no changes, leave it unchanged.