Journal Tier Strategy

Three-tier framing strategy for cover letters. Each template references its tier; the writing rules below apply per tier and override venue-specific style when not contradicted.

top-journal

Nature family, Science family, Cell Press.

• Opening: lead with the scientific advance, not the topic. The first sentence must say what changes in the field because of this work.
• Novelty framing: paradigm-shift level. "We resolve the long-standing discrepancy between X and Y" beats "We propose a new method for X."
• Word budget: ≤350 words. Editors expect tightness.
• Significance threshold: must matter for a broad scientific audience, not a subfield. If it only matters to a subfield, route to a tier-2 sister journal.
• Quantitative anchor: required. At least one headline number traceable to the manuscript.
• Comparison frame: cite a recent paper from the same journal that this work directly extends or contradicts. Editors notice this and it signals real reading of the venue.
• Cliché avoidance: never open with "We are pleased to submit..." — flagged as low-effort.

mid-journal

IEEE Transactions, ACM journals, Springer LNCS / LNAI, PLOS family, Elsevier specialized journals.

• Opening: methodological contribution stated cleanly. "We introduce X, an algorithm that handles Y by Z."
• Novelty framing: contribution-led. Enumerate 3-4 specific contributions.
• Word budget: 400-500 words. More room for methodological context.
• Significance threshold: matters within the venue's specialty area. Broader-impact framing is optional; deep methodological framing is required.
• Quantitative anchor: required, with explicit comparator naming (baseline + dataset + protocol).
• Comparison frame: cite one or two recent papers from the same journal to position the work.
• Conference extension: if extending a prior conference paper, disclose the venue and percentage of new content (IEEE / ACM typical bar: ≥30%).

conference

NeurIPS, ICML, CVPR, ICCV, ECCV, ACL, EMNLP, AAAI, IJCAI, KDD, WWW.

The major ML/vision conferences do not use a cover letter. NeurIPS, ICML, CVPR, ICLR and peers collect submission metadata through structured OpenReview / CMT forms (author list, abstract, checklists), not a free-form letter. Treat the neurips / icml / cvpr templates as applicable only when a workshop, datasets-and-benchmarks track, or special call explicitly requests a letter. For the main track, redirect the user to the submission form rather than generating a letter.

• Opening: technical contribution stated in one sentence.
• Novelty framing: contribution-led, concise.
• Word budget: ≤400 words. Conference reviewers do not read longer letters carefully.
• Significance threshold: contributes to a specific subfield with strong empirical or theoretical evidence.
• Quantitative anchor: required, with dataset and split named.
• No broader-impact rhetoric in the letter itself: that belongs in the manuscript's dedicated Broader Impact section (many ML venues now require it).
• Dual-submission disclosure: conferences are strict; always confirm compliance.
• Anonymization: for double-blind venues, the cover letter is typically not anonymized but should not leak identifying information beyond what is already in the submission system.

When the venue does not match any tier

Fall back to templates/generic.md. Manually pick the closest tier for framing guidance. If unsure between tiers, prefer the more conservative (mid-journal) framing — overclaiming is worse than underclaiming for editorial trust.

Editorial practice snapshot (2025-26)

The cover letter is the first step of desk review (high-volume journals desk-reject 50-80%), so these venue-specific realities matter:

• AI-use disclosure is now a cover-letter requirement at the top tier. ICMJE (Jan 2026, Section V) requires generative-AI use to be described in the cover letter and the manuscript; Science/AAAS, NEJM, and APS require cover-letter disclosure explicitly. IEEE (Acknowledgments), ACM (within the Work), and Elsevier/Springer (Methods / dedicated statement) place it in the manuscript instead. Undisclosed AI use can be treated as misconduct.
• Nature-family cover letters are a confidential channel — not shown to reviewers. Use them for competing interests, related work under consideration, or specific editorial-handling requests; do not restate the abstract. The letter is optional for Nature flagship.
• IEEE is bimodal: Proceedings of the IEEE mandates a detailed cover letter (a missing one can be returned); IEEE TIM explicitly says not to submit one unless there is something specific to raise. Always check the target title's author center.
• Author-suggested reviewers are being curtailed (Cureus removed the option in 2025; PLOS ONE removed it earlier; Wiley/IOP now require independent verification). Do not generate a suggested-reviewers paragraph by default — include it only when the target venue's submission process asks for it.
• Rejected-and-resubmitted manuscripts: some venues (Wiley Small, IEEE/ACM ToN) require the cover letter to state the prior submission and summarize the changes.
• Double-blind venues: list people who have already seen the manuscript (to protect the blind), and disclose overlap with the authors' own prior work or shared datasets.
• Springer Nature fixed phrasing: "We confirm that this manuscript has not been published elsewhere and is not under consideration by another journal. All authors have approved the manuscript and agree with its submission to [journal]."
• Generic template letters are a negative signal: editors who screen 50+ letters a week spot mass-mailed wording instantly. Specificity (named scope dimension, a recent same-venue paper, the concrete headline result) is what reads as written-with-attention.