Skip to main content

High-impact AI in editorial review and decision making

The guidance below highlights impactful applications of AI during editorial review and decision making, with practical considerations and prompt templates you can copy and paste to adapt to your own needs.

Aim: Use AI to strengthen structure, aid constructive discussion and enhance quality assessment - not to outsource judgement.

Woman, doctor and laptop in office at night at workplace, hospital or desk for medical career vision. Healthcare expert, dark clinic or focus for reading at computer, email or cancer research article

High-impact use

  • Create a review feedback structure template (major/minor issues, methods, stats, reporting, ethics, novelty, clarity) so feedback is consistent and complete.

  • Generate ‘questions to ask’ to help form your own critique e.g. missing controls, unclear inclusion criteria, under-described methods, inconsistent numbers, weak causal language.

  • Run consistency and completeness checks: does the abstract match results; do figures match text; are key limitations stated; are claims calibrated to evidence?

  • Draft clearer decision letters that separate fatal vs fixable issues, point to exact locations, and request specific revisions/tests.

  • Bias guardrails: prompt for alternative interpretations and ‘what would change my mind?’ checks before recommending a decision.

Safety check

  • Confidentiality is non-negotiable: don’t paste manuscripts, figures, reviewer reports, or decision discussions into tools that are unauthorized for that data. If your policy/tooling isn’t explicitly closed and permitted for confidential content, don’t use it.

  • Humans remain accountable: AI can support process but must not be the decision-maker or the source of scientific/ethical judgement.

  • Avoid hallucinated critique: verify every AI-suggested concern against the manuscript (methods, data, claims). If you can’t point to where it appears, don’t include it.

  • No ‘synthetic authority’: don’t let AI invent standards, policies, or ‘best practices’ claims—anchor feedback to journal requirements, reporting guidelines, and what’s actually in the paper.

  • Don’t generate accusations: if something looks suspicious (images, authorship, provenance), phrase it as a request for clarification/evidence and escalate through the journal’s integrity process.

  • Disclosure & record-keeping: follow the journal/publisher policy on whether reviewers/editors must declare AI assistance; keep a brief internal note of how AI was used (e.g. ‘structure only,’ ‘consistency checklist’,’ ‘language clarity’).

  • Check journal policies: adhere to relevant policies from the journal or publisher.

Copy and paste prompts

CONFIDENTIALITY: Only use the following prompts with tools explicitly approved for confidential content. Do not paste manuscript text, reviewer reports, or editorial correspondence into general-purpose AI tools

📑 Copy and paste prompt: internal consistency checker

Run an internal consistency check on the manuscript text I paste. Find mismatches or contradictions across: abstract, methods, results, tables/figures, and conclusions.

Output:

- Issue (what doesn’t match)

- Where it appears (exact quote/section)

- Why it matters

- A neutral question/request to the authors

Rules: No new claims. If you can’t locate it, don’t include it.

📑 Copy and paste prompt: claims calibration

Help me check whether claims are calibrated to evidence. From the pasted Results/Discussion, extract the main claims and tag each as: Supported / Overstated / Not supported by provided text. For each, cite the sentence(s) and suggest safer wording or a question to authors.

Rule: Do not infer causation unless the design described supports it.

📑 Copy and paste prompt: decision letter drafter (editor)

Draft a clear decision letter based on my notes (below). Manuscript ID: [ID] Decision category: [Major revision / Minor revision / Reject but encourage resubmission] My bullet notes (paste): [notes]

Output a letter that:

- briefly summarizes the key reason(s)

- separates fatal issues vs fixable issues

- requests specific revisions/tests and points to exact sections

- keeps tone constructive and neutral

Rules: Don’t add new critique beyond my notes. Don’t invent policy. No accusations—phrase concerns as requests for clarification/evidence.

📑 Copy and paste prompt: reporting guideline spot-check

Identify the likely reporting checklist for this study type (e.g., CONSORT/STROBE/PRISMA/etc.) and produce a short ‘spot-check’ list of the most commonly missing items. Then, for each item, tell me whether it appears in the pasted text and where (quote/section), or mark Missing.

Rules: Don’t invent items not in standard guidelines; if unsure, say so.

Explore all research stages