A graphical abstract is one image that summarises a paper's main finding for a reader scanning a contents list. Build it from your text abstract: the question, the design and population, the main result with its effect size and confidence interval, and the key limitation. Lay it out to read in one direction, and use the paper's own numbers.
Journals use the names loosely: graphical abstract, visual abstract, infographic, or a table of contents (TOC) graphic in some chemistry journals. This is general guidance, not a Directive Publications requirement; your target journal's instructions come first.
What a graphical abstract has to carry, panel by panel
A graphical abstract's content has its own reporting guideline. RIVA-C (Reporting Infographics and Visual Abstracts of Comparative studies) is a 10-item checklist published in 2024 and listed in the EQUATOR Network library. It covers content, not design, and only comparative studies of interventions, such as randomised trials, retrospective observational and pre-post cohort studies, and systematic reviews. Case reports, case series, cross-sectional, diagnostic and prognostic studies are outside its scope, though its principles still help.
Two RIVA-C points are easy to get wrong. Show the difference between groups with its precision, not each group's change from baseline, which gives a biased estimate of the effect. Prefer absolute effects, because relative effects can make an effect look much larger; showing both is acceptable. The International Committee of Medical Journal Editors (ICMJE) already asks that the text abstract give effect sizes and note important limitations without overinterpreting the findings.
Copy the first two columns of the table as your template. The example below is invented to illustrate the structure; its numbers are not real data.
| Panel (RIVA-C items) | What it carries | Invented example |
|---|---|---|
| Headline | The question or finding, in the paper's wording; no cliff-hanger | Structured discharge checklist or usual discharge: 30-day readmissions |
| Who and how (items 1–2) | Design, setting, sample size, key characteristics | Randomised trial, 2 hospitals, 400 adults discharged from medical wards |
| Comparison (item 3) | Intervention and comparator | Checklist (n = 200) vs usual discharge (n = 200) |
| Primary outcome (item 4) | Labelled primary, with scale, units and time point | Primary outcome: readmission to the same hospital within 30 days, % of patients |
| Main result (items 5–6) | Between-group effect with its 95% confidence interval (CI), absolute where possible; numbers analysed per group; any justified threshold for clinical importance | 24/196 (12.2%) vs 39/194 (20.1%); difference −7.9 percentage points (95% CI −15.1 to −0.6) |
| Harms (item 7) | Serious adverse events per group, or a note that they were not measured or not reported | Serious adverse events: 5/196 vs 6/194 |
| Certainty (item 8) | Certainty of evidence; systematic reviews only | Not applicable |
| Take-away (items 9–10) and limitation | A conclusion matched to the population, comparison and primary outcome that acknowledges harms; the key limitation | Fewer 30-day readmissions with the checklist than with usual discharge; serious adverse events too few to compare; wide interval; readmissions to other hospitals not counted |
| Route to the paper | Citation, DOI or link | [first author] et al. [journal] [year]. doi.org/10.xxxx/example |
Lay out a graphical abstract that reads in one direction
Elsevier asks for a clear start and end, preferably reading top to bottom or left to right, and Cell Press agrees on direction. Two layouts suit a two-group comparison:
- Three panels, left to right: who was studied, what was compared, what was found, with the headline above.
- Two columns, one per group: each with its icon, denominator and result, and the between-group difference with its interval in a strip beneath.
Treat words as a budget: a label, a number and a short phrase per panel. Elsevier shrinks its graphical abstract to fit a contents-list window 200 pixels high, so check every label at that height or on a phone. The Web Content Accessibility Guidelines apply contrast thresholds to images of text too. Palettes, contrast ratios and alt text for multi-panel figures are covered in making figures and text accessible to every reader. Direct labels, colour and honest axes are covered in preparing figures and tables for publication.
Publishers disagree about numbers in the graphic. Cell Press's 2018 guidance said to include no data items, while visual abstracts report outcome numbers with their units. If your journal bans data, keep the numbers in the text abstract and the picture within what they support.
Take the graphical abstract's size and file type from the target journal
No single pixel size, shape or file type applies everywhere. These examples were checked in October 2026 and can change:
- Elsevier: a wide image of at least 1328 × 531 pixels at a minimum of 300 dots per inch (dpi), in a 500:200 ratio; Microsoft Office files are among its preferred types.
- Cell Press: an exact square of 1200 × 1200 pixels at 300 dpi, with TIFF, PDF or JPG as its preferred types.
- American Chemical Society (ACS): 3.25 × 1.75 inches for The Journal of Physical Chemistry Letters, but 2 inches high by 3 wide for ACS Energy Letters.
Set the canvas to the journal's size before you draw, because squeezing a wide layout into a square breaks the reading order. Check, too, whether the graphic goes in the manuscript file or as a separate upload.
Use only icons and images you have the right to publish
Every icon you did not draw yourself carries a licence, and the Creative Commons FAQ says the original licence still applies after you adapt it. Check each icon against your article's licence.
| Icon's licence | Fits a CC BY article? | Why |
|---|---|---|
| Public domain or CC0 | Yes | No credit required, though some libraries ask for it |
| CC BY | Yes | Credit the creator, link the licence and note changes |
| CC BY-SA | Ask the journal first | Adaptations must carry the same licence |
| CC BY-NC | No | CC BY allows the commercial reuse that NonCommercial material forbids |
| CC BY-ND | No | NoDerivatives forbids sharing adaptations, which CC BY lets anyone make of your article |
| Subscription illustration tool | Only on your plan's terms | Some grant publication rights only on paid plans, require a credit line and keep ownership of their artwork |
In a free library, the licence can differ from one entry to the next. The National Institutes of Health (NIH) BioArt Source labels some entries public domain and others CC BY, so check each icon, not the library. Credit each icon to its own source, worded as in how to attribute reused Creative Commons material. Use no journal's or other organisation's logo or brand without permission. Elsevier bars general-purpose generative artificial intelligence (AI) image tools from graphical abstracts, so check your journal's policy before using one. To reuse your own graphical abstract in slides or social media, check what you signed: CC BY allows reuse with credit, but a copyright transfer or exclusive licence can limit it.
Keep the picture no bolder than the paper
In 129 journal infographics, risk of bias was acknowledged in 2%, and none of the 69 studies that declared a conflict of interest disclosed it in the infographic (Ferreira and colleagues, 2022). A third of infographics for randomised trials with non-significant results contained spin, against 26% of the same trials' abstracts, not a significant difference (Muller and colleagues, 2025).
- No causal arrows for an association. An arrow from exposure to outcome reads as "causes", so set observational groups side by side. The wording ladder that matches claims to study design governs pictures as much as verbs.
- The paper's denominators, printed. Array size changes perceived risk: people shown 1,000 icons rather than 100 judged risks more serious and risk reductions larger (Galesic and colleagues, 2009). Use the paper's denominator and state it.
- No implied efficacy without a comparison group. Before-and-after panels from a single-arm series look like a treatment effect.
- Uncertainty on show. Print the interval beside every estimate, and draw no equals sign for a non-significant difference, because its interval can still include an effect that matters.
- Honest scale. A pictogram enlarged in both dimensions grows by area and exaggerates the difference.
- Primary outcome first, not a striking secondary result.
- No identifiable patients. The ICMJE allows identifying photographs only when essential for scientific purposes and with written informed consent for publication from the patient (or parent or guardian). Consent and file cleaning are covered in patient consent for case reports and clinical photographs and de-identifying clinical images and scans. A generic drawing, not traced from a patient's photograph, avoids the problem.
Graphical abstract checklist before you upload
- Size, shape and file type match the journal's current instructions.
- Every number and denominator matches the results and the text abstract.
- Design, sample size, comparator and the labelled primary outcome are shown.
- The main result is a between-group effect with its interval, absolute where possible.
- Harms are shown, or the graphic says they were not measured or not reported.
- The take-away is no stronger than the paper's conclusion and names the key limitation.
- It reads in one direction and is legible at display size.
- A colleague who has not read the paper can state the question, the main result and its limit from the graphic alone.
- No arrow implies cause for an association, and no single-arm result is drawn as an effect.
- Every icon's licence fits the article's, with credits given.
- No patient is identifiable, and no logo or brand appears without permission.
- Any AI use follows the journal's policy.
- A citation, DOI or link leads to the paper.
Our author guidelines do not ask for a graphical abstract
Directive Publications' publisher-wide author guidelines do not ask for a graphical abstract, and none of our policy pages mention one. If your target journal asks for one, follow its instructions. If you send us a manuscript, our own rules apply to every figure and image in it. Articles are published under CC BY 4.0 and authors keep copyright, so third-party material, icons included, must be compatible with that licence, as our Copyright, Licensing & Author Rights page sets out. Our author guidelines require separate, explicit consent for publication of any identifiable image, and our editorial policies require any use of AI for image generation to be disclosed. If a figure in one of our articles misstates its paper, report the problem to us.