The Publication-Ready Figures Checklist (Save This Before Your Next Submission)

Desk rejections for figure problems are the most preventable failures in publishing. Not weak data, not contested interpretations — just axis labels at 4 pt, a 72 dpi export, an error bar nobody defined. This checklist exists so those never happen to you. It is organized by workflow stage and written to be printed or pasted beside your submission todo list.

Run it twice: once when you think the figures are done, and once the day of submission, because the version you submit is sometimes not the version you checked.

Content accuracy

• Every figure makes its one-sentence message obvious within seconds
• Axis baselines are honest (bar charts start at zero — non-negotiable)
• Error bars are present where means appear, and the caption states what they represent (SD / SEM / CI)
• Sample sizes (n) are visible on the figure or in the caption for every condition
• Panel letters (a, b, c…) follow the journal's convention and match the caption order exactly
• Every label, abbreviation, and unit is spelled correctly — read labels aloud to catch what your eyes slide past
• Schematic content is scientifically verified against your source material (organelles inside cells, arrows in physically meaningful directions)
• If any illustrative element was AI-assisted, you have verified every scientific relationship in it yourself

Format, resolution, and files

• Each figure exported in a format the journal accepts (vector PDF/SVG/EPS for plots and schematics; high-res raster for imagery)
• Resolution meets the journal's stated minimum for raster images — check current author guidelines, they differ between publishers
• Dimensions match the intended column width (single or double) at final print size
• Fonts are embedded in the exported file
• Files named systematically: Fig1.tiff, Fig2.pdf — no final_v3_REAL archaeology for the production team
• Source files archived in your lab storage; the PDF you submit is not the only copy in existence

Typography and readability

• All text legible at final print size (shrink to column width and read it; if you squint, it is too small)
• One font family, hierarchy by weight/size only
• Axis titles include units in parentheses: "Time (min)"
• Tick labels upright or rotated consistently; no overlapping labels
• Direct labels used wherever a boxed legend can be replaced

Color and accessibility

• One fixed color per condition, consistent across all figures in the manuscript
• Colorblind-safe palette — no red/green as the only contrast carrying meaning
• Meaning survives greyscale printing (shape, line style, or labels redundantly encode)
• The finding carries the strongest contrast in the figure; decorations step down

Captions

• Caption states the message, not just the mechanics — first sentence carries the finding
• Every symbol, line style, and abbreviation in the figure is defined in the caption
• Statistical notation explained: what the error bars are, what the asterisks mean and against what threshold
• Panel letters referenced in order
• Captions spell-checked separately — captions are typed last and it shows

Disclosure and ethics

• AI-use disclosure prepared if generative AI contributed to figures or text — most major publishers now expect this in methods or acknowledgments; check your specific journal's policy
• No figure previously published elsewhere without permission and citation
• Adapted figures credited per the journal's reuse requirements
• Image adjustments limited to what the journal allows (typically global brightness/contrast, never feature-altering edits, applied identically across compared groups)

The 15-minute final pass

The day of submission, with the compiled manuscript PDF open:

1. Read only the figures and captions in order — does the paper's story hold up from figures alone?
2. Zoom to 100% on every figure page: any blurring, font corruption, or color shifts from the compilation?
3. Search the PDF for each figure's filename to confirm the right file versions made it in
4. Confirm the figure list in the submission portal matches the manuscript, letter for letter
5. Archive the exact submitted files in your records — future you, during revisions, will need the identical sources

For the design principles behind most of these checks — hierarchy, honest baselines, label placement — our walkthrough of scientific illustration for research papers explains the reasoning. And if the checklist exposed a graphical abstract that needs rebuilding, the graphical abstract template guide has you covered.

Key takeaways

• Most desk-rejection figure problems are procedural — resolution, fonts, labels — and fully preventable with a checklist
• Run the checklist twice: at "done" and again the day of submission
• Captions deserve the same rigor as figures: message first, every symbol defined, statistics explicit
• AI-assisted figures add a new line to the ethics section: verify content yourself and disclose per journal policy

Frequently asked questions

What resolution do journals actually require for figures? Requirements vary by publisher — commonly 300 dpi for color/halftone images and 600–1000 dpi for line art — but the current numbers live in each journal's author guidelines and they do change. The practical rule: export vectors wherever accepted, and when raster is required, follow the journal's stated dpi at final print dimensions.

Can I fix figure problems at the revision stage instead? Sometimes, but you are gambling. Formatting failures can trigger desk rejection before scientific review begins, and rushed fixes at revision collide with a new deadline. The checklist costs twenty minutes; a restart of the submission clock costs weeks.

Do these checks apply to preprints? The accuracy and accessibility checks do — a preprint with unreadable figures damages the paper's reception in your field. Format and resolution checks matter less on most preprint servers but become critical the moment you submit to a journal, so running the full list at preprint stage puts you ahead of the submission rush.

How do I keep figure formatting consistent across co-authors' contributions? Share a single template: the palette file, the font, and one exemplar figure with the caption structure. One person owns final assembly — distributed assembly is how manuscripts end up with three different sans-serifs and two conflicting color codes.

Figures that pass this checklist deserve to be seen. SciScroll's discover feed is built exactly for showing research as figures — share yours, browse others', and generate new ones with AI assistance (free to start; plans for heavier use).