Free scoping read. A written assessment of your manuscript in two business days, at no cost.

Start one
Contents

Academy · Graphical abstract · How to make one

Making a Graphical Abstract for a Research Paper: Publisher Specifications, Permissions and Accessibility

Four publishers have four different specifications, and two of these publishers belong to the same company. Authors should therefore consult the target journal’s own requirements before preparing a graphical abstract.

Pubrica AcademyWritten by a research writing specialist Reviewed bya research methodology reviewer Published15 October 2026 Last checked15 October 2026 Reading time15 minutes Sources11 primary

A graphical abstract is a single visual summary of the principal message of a paper, displayed in search results and tables of contents rather than in the article itself. The central task in preparing one is identifying the specification of the target journal, because specifications differ considerably, including between imprints of the same publisher.

Definition of a graphical abstract

Elsevier's description: “A graphical abstract summarizes the contents of your article in a concise, pictorial, professional form”1, intended to “capture the content of the article for readers at a single glance”1 Note where it appears: “The graphical abstract will be displayed in online search result lists”1 and “will typically not appear in the article PDF file or print.”1 It is a discovery device, which is why legibility at small sizes governs every decision about it.

A note on terminology

The terminology warrants comment, because there is no settled taxonomy and publishers themselves are inconsistent. Elsevier uses both terms in a single sentence — journals are “increasingly requesting the submission of a 'graphical' or 'visual abstract'” — and elsewhere describes a visual abstract as commonly a PowerPoint slide carrying title, methods, results, conclusions and the citation, which is close to the opposite of its own graphical abstract specification (“no additional text, outline, or synopsis should be included”). Cell Press will not accept PowerPoint files. Every non-publisher source identified that distinguishes graphical abstracts, visual abstracts and infographics was a design or editing vendor, and none is therefore cited here. The labels should be treated as unstandardised, and authors should work from the target journal's own definition.

Whether a graphical abstract is required

Nor is there a general rule on whether one is required. One Elsevier journal guideline is headed “Guidelines for graphical abstracts (optional)”; Cell Press says “most Cell Press journals will ask for a Graphical Abstract”; at least one journal requires them for all accepted papers. Elsevier's own advice is to review the journal's guide for authors, which is the only reliable general statement available.

Conflicting publisher specifications

Consulting the journal’s specification first

Before any design work begins, the target journal's own specification should be identified, because these requirements conflict, and not only between publishers. Elsevier asks for a 2.5 : 1 landscape image; Cell Press, which Elsevier owns, requires a 1 : 1 square single panel. An image built to one cannot be submitted to the other. Elsevier names Microsoft PowerPoint as its template format; Cell Press does not accept PowerPoint files at all. MDPI, moreover, states its dimensions as height by width, the reverse of Elsevier's convention, so “560 × 1100” is a portrait minimum rather than a landscape one. There is no industry standard here, and every publisher consulted advises checking the journal's own guide.

Table 1. Four publishers’ graphical abstract specifications

ElsevierCell PressMDPIIOP Publishing
Minimum dimensions1328 × 531 px (w × h)1200 × 1200 px560 × 1100 px (h × w)None specified
Aspect ratio500 : 200, i.e. 2.5 : 1 landscape1 : 1 squareKeep ratio, do not distort—
ResolutionMin 300 dpi300 dpiNot statedNot stated
File formatsTIFF, EPS, PDF, MS OfficeTIFF, PDF, JPG — no PowerPointJPEG, PNG, TIFF, SVGNot stated
FontsTimes, Arial, Courier or SymbolArial 12–16 ptClear and professionalNot stated
Single panel requiredNot statedYesNot statedNot stated
Reuse a figure from the paperPermitted in general guidance; a journal-level guideline says neverData items prohibitedProhibitedPermitted, but no graphs

The underlying quotes, so you can check them. Elsevier: “provide an image with a minimum of 1328 x 531 pixels (w x h)”1 “using a minimum resolution of 300 dpi”1, and “please use the same ratio (500 wide x 200 high)”1 “preferred file types are TIFF, EPS, PDF or MS Office files”1 and for type, “please use Times, Arial, Courier or Symbol”1 Cell Press: “A one panel 1200 pixel square at 300 dpi.”2 “Square, equal on all sides.”2 with type set in “Arial, 12–16 points. Smaller fonts will not be legible online.”2 MDPI: “the minimum size is 560 × 1100 pixels, with a maximum size of 2800 × 5500 pixels (height × width)”3 IOP, differing from all of them: authors “do not need to prepare the image file to a particular size.”4

On text. Elsevier is the strictest: “No additional text, outline, or synopsis should be included.”1 and “Any text or label must be part of the image file.”1 MDPI: “Text should only be used for labels and graphs; avoid prose.”3 Cell Press asks that “Use simple labels and add text sparingly”2, with the test being that “readers should be able to understand the image without needing a dictionary.”2

On empty space, which Cell Press names as the commonest author mistake: MDPI requires that “Any blank space surrounding the GA must be removed.”3

Reuse of figures from the paper

A journal-level question, not a principle

One rule that is often stated as absolute is not: whether a graphical abstract may reuse a figure from the paper. MDPI prohibits it. IOP explicitly permits it — “this image does not have to be a figure from your article” — while barring graphs. And Elsevier's general guidance contemplates that “some authors may use a concluding figure from the article”, while one of its own journal-level guidelines says a graphical abstract “should never copy a figure or table from the paper”. This is therefore a journal-level question, not a principle.

IOP's wording: “This image does not have to be a figure from your article.”4 — though “we ask that you do not submit images of/from graphs.”4 Cell Press's prohibition appears in its list of things not to do: “Include data items of any type; all the content should be in a graphical form.”2 And Elsevier requires in all cases that “Authors must provide an original image that clearly represents the work described in the paper.”1

Preparation of a graphical abstract

Table 2. Ten steps

Find your journal’s specification firstThe target journal’s specification should be used, not a general guide or another journal’s. The dimensions, the aspect ratio, the formats and the single-panel rule all differ, and an image built to the wrong specification is rejected or must be rebuilt.
Decide the one messageA graphical abstract carries the take-home message, not the study. If the message cannot be stated in a single sentence, the image will not convey it either.
Sketch the reading orderElsevier asks for “a clear start and end” reading top to bottom or left to right. The reading order should be decided before any drawing begins.
Check whether you may reuse a figureMDPI prohibits it, IOP permits it, and Elsevier’s own guidance is inconsistent between levels. This is a journal-level question.
Build it at the required pixel size and resolutionThe image should not be scaled up afterwards. Resolution should be 300 dpi where specified, and text should form part of the image file.
Set type in a permitted face at a legible sizeCell Press asks for Arial at 12 to 16 points; Elsevier names Times, Arial, Courier or Symbol. Legibility should be checked at the size at which the image will be displayed.
Strip everything that is not carrying meaningExcess white space is the mistake Cell Press names as most common; MDPI requires blank space around the image to be removed.
Clear the rights on every elementIOP requires an image that needs no permission at all; Elsevier requires an original image and permission for third-party material.
Do not generate it with AIElsevier prohibits AI-created or AI-altered images and names graphical abstracts explicitly.
Write the alt textAlt text should be under 100 words, generally 25 to 30, and convey the same meaning as the image. It should be written at submission: at least one publisher will not correct alt text after publication.

AI-generated images

A named prohibition

This point is relevant before any image generator is used. Elsevier explicitly prohibits AI-generated graphical abstracts, by name. Its policy states that Elsevier “does not permit the use of generative AI or AI-assisted tools to create or alter images” in submitted manuscripts, and that using them for artwork “such as for book or commissioned content covers or graphical abstracts is not permitted”. Adjustments to brightness, contrast and colour balance remain acceptable provided they do not obscure information, and AI that is part of the research method is a separate case that must be described reproducibly. The policy also notes that image forensics tools may be applied to submissions. This was verified for Elsevier; equivalent policies at other publishers could not be accessed, and authors should therefore check their own publisher’s policy.

The policy's own words: “Elsevier does not permit the use of generative AI or AI-assisted tools to create or alter images”5 in submitted manuscripts, and use for artwork “such as for book or commissioned content covers or graphical abstracts is not permitted.”5 It also notes that image forensics tools “might be applied to submitted manuscripts”5

Rights and permissions

Three different standards, of which IOP's is the strictest and simplest. Elsevier: “Please ensure that you have obtained the necessary permission”1 for third-party material. IOP: the image “should not be under any Copyright, or require any permission for its reuse.”4 — that is, not “get permission” but “use something that needs none”. MDPI prohibits an image taken from elsewhere without the necessary rights3. Icons, photographs and figure elements all fall under this, and it is worth settling before you build rather than after.

Alt text

A frequently omitted step

This step is omitted entirely by most graphical-abstract guides. A graphical abstract carries the paper's message as a picture, which means that without a text alternative the message is unavailable to anyone using a screen reader and inaccessible to any system that reads text. WCAG's requirement is unambiguous: all non-text content “has a text alternative that serves the equivalent purpose”, and a graphical abstract falls under none of the criterion's exceptions. Springer Nature states that it “expect[s] all images in manuscripts to have sufficient alt text” which “must convey the same meaning as the image”, and, of relevance at submission, that it does not make post-publication corrections to alt text. Taylor & Francis gives the only concrete length target identified: under 100 words, and generally 25 to 30.

WCAG's normative requirement: “All non-text content that is presented to the user has a text alternative that serves the equivalent purpose, except for the situations listed below.”6 Springer Nature: “We expect all images in manuscripts to have sufficient alt text.”7, which “must convey the same meaning as the image.”7, and “Authors remain solely responsible for the accuracy of all alt text associated with their manuscripts.”7 Note the timing point: it “does not make post-publication corrections to alt text”7, so this is a submission-stage task. Taylor & Francis gives the length: “Alt text should strive to be under 100 words and generally 25 to 30 words long.”8 and “Your alt text should be included along with the figure captions.”8

One qualification is necessary: none of the publisher accessibility pages consulted mentions graphical abstracts by name. Their guidance is about figures and images generally, and applying it to a graphical abstract is an inference made in this article; it is a straightforward inference, since a graphical abstract is an image, but an inference rather than a quoted rule.

Evidence of effectiveness

The extent of the evidence

The evidence supports a narrower conclusion than is usually claimed, although that conclusion still provides a reason to prepare one. The well-known study found large effects on social-media metrics at one journal in 2016. The independent replication found a much smaller effect (roughly twice the impressions rather than nearly eight times) and, importantly, no significant difference in link clicks (p = .35), which is the outcome most closely related to reading the paper. A separate retrospective analysis in sport science found articles with graphical or video abstracts had lower altmetric scores and fewer citations, although, as the study was observational, this may reflect which articles received them. Three randomised trials found that an infographic produced no better knowledge than a plain-language summary, despite readers preferring it. In summary, there is good evidence for short-term visibility on social platforms but no good evidence for effects on comprehension, downloads or citations.

The original study, 44 articles each tweeted twice with a washout period9:

Table 3. Title alone versus title plus visual abstract

OutcomeTitle aloneWith visual abstractEffectp
Impressions3,073.323,611.27.7-fold<0.001
Retweets11.092.18.4-fold<0.001
Article visits65.6175.42.7-fold<0.001

The independent replication, 40 articles each tweeted three ways10:

Table 4. Replication: citation only, key figure, or visual abstract

OutcomeCitation onlyKey figureVisual abstractp
Impressions638.58731.881,351.08<.001
Total engagements15.7524.0883.75<.001
Retweets0.951.303.68<.001
Link clicks8.136.688.95.35

Note the final row. The replication's authors conclude that “The use of visual abstracts increased visibility of research articles on Twitter”10 — visibility, specifically. On the measure closest to someone reading the paper, link clicks, the difference was not significant10.

And on comprehension, three randomised trials comparing an infographic with a plain-language summary: “We found no difference in knowledge between the infographic and the text-based PLS in any of the trials.”11 Readers preferred the infographic and rated it better to read; they did not learn more from it11.

One figure is not relied upon here: Elsevier's author page states that average annual article use is doubled where there is a visual abstract, supported by a hyperlink to a paper on a host closed to automated retrieval. Because that paper could not be accessed, its sample, period and method cannot be reported, and the claim is therefore not endorsed here.

Pre-submission checks

Practical application

  1. The guide for authors of the target journal should be consulted and its specifications followed, rather than those of a general guide, because specifications conflict between publishers.
  2. The aspect ratio should be checked specifically. A 2.5:1 landscape format and a 1:1 square are not interchangeable, and rebuilding an image requires more work than preparing it correctly at the outset.
  3. MDPI-style dimensions should be read carefully, because MDPI states height by width, the reverse of the usual convention.
  4. The image should be built at the specified pixel size and resolution rather than scaled up.
  5. Legibility should be checked at display size rather than at full size, because graphical abstracts are displayed small in search results.
  6. White space and any element that does not carry meaning should be removed.
  7. Authors should confirm whether the target journal permits reuse of a figure from the paper, as this varies.
  8. The image should not be generated or altered with AI where the publisher prohibits it; Elsevier names graphical abstracts explicitly.
  9. Permissions should be cleared for every element, or elements requiring no permission should be used.
  10. Alt text should be written and submitted with the figure. It should be under 100 words, convey the same meaning as the image, and be prepared at submission rather than after publication.

Key points

The target journal’s own specification should be consulted before preparation: Elsevier requires 1328 × 531 at 2.5:1, Cell Press a 1200-pixel single-panel square, MDPI states its range as height by width, and IOP sets no size at all. Whether a figure from the paper may be reused varies by journal. Elsevier prohibits AI-generated graphical abstracts by name. Alt text should be written at submission, because at least one publisher will not correct it afterwards. The evidence supports short-term social-media visibility only; the replication found no significant effect on link-throughs, and no study has shown a citation benefit.

It depends entirely on the journal, and the requirements conflict. Elsevier asks for a minimum of 1328 × 531 pixels at 300 dpi in a 2.5:1 landscape ratio. Cell Press — which Elsevier owns — requires “a one panel 1200 pixel square at 300 dpi”. An image built for one cannot be submitted to the other, so the target journal’s own specification should be identified first.

For Cell Press, yes — “a single-panel image”. Elsevier states no panel rule at all; it asks instead for “a clear start and end” reading top to bottom or left to right. So the single-panel rule is a Cell Press requirement, not a general one, and should not be attributed to publishers at large.

It varies, although it is often stated as an absolute rule. MDPI prohibits it. IOP Publishing explicitly permits it — “this image does not have to be a figure from your article” — while barring graphs. Elsevier’s general guidance contemplates using a concluding figure, while one of its own journal-level guidelines says never to copy a figure or table.

Not for Elsevier. Its policy states that Elsevier “does not permit the use of generative AI or AI-assisted tools to create or alter images” in submitted manuscripts, and that use for artwork “such as for book or commissioned content covers or graphical abstracts is not permitted”. It also notes that image forensics tools may be applied. Authors should check their own publisher’s policy.

Only as much as is needed to carry the meaning. Elsevier states “no additional text, outline, or synopsis should be included” and that any label must be part of the image file. MDPI states that “text should only be used for labels and graphs; avoid prose”, and warns that images with excessive text will be rejected. Cell Press asks for simple labels used sparingly.

There is no good evidence that they do. The well-known studies measured social-media metrics: the original found large effects on impressions and retweets, but the independent replication found no significant difference in link clicks (p = .35). One retrospective analysis in sport science found articles with graphical abstracts had fewer citations. There is evidence for visibility, but none for citations.

Yes. WCAG requires that all non-text content “has a text alternative that serves the equivalent purpose”, and a graphical abstract meets none of the exceptions. Springer Nature “expect[s] all images in manuscripts to have sufficient alt text” which “must convey the same meaning as the image”. Taylor & Francis suggests under 100 words, generally 25 to 30.

The terms are not standardised, and publishers use them inconsistently. Elsevier uses both in one sentence, and elsewhere describes a visual abstract as commonly a PowerPoint slide carrying title, methods, results and the citation — close to the opposite of its own graphical abstract specification. Authors should work from the target journal’s definition.

References

  1. Elsevier. Graphical abstract. Author resources. read 15 October 2026. View on ElsevierThe originating publisher's specification. Note it states no single-panel rule and gives no millimetre size, contrary to what is often claimed of it.
  2. Cell Press. Attract readers at a glance with your graphical abstract. CrossTalk. read 15 October 2026. View on Cell PressCell Press's own guidance, and the source of the single-panel and square requirements. It is a publisher blog post rather than the author PDF, which we could not reach.
  3. MDPI. Maximising the impact of your graphical abstracts. MDPI blog, 9 November 2021. read 15 October 2026. View on MDPINote MDPI states its pixel range as height by width, the reverse of Elsevier's convention. Transposing it is the easiest mistake to make on this subject.
  4. IOP Publishing. Graphical abstracts. Publishing support. read 15 October 2026. View on IOP PublishingIncluded because it differs from the others in both directions: no size requirement at all, and it permits a figure from the article, which MDPI and others prohibit.
  5. Elsevier. The use of generative AI and AI-assisted technologies in writing for Elsevier. Policies and standards. read 15 October 2026. View on ElsevierA named prohibition on AI-generated graphical abstracts. We verified this for Elsevier only, so it is not presented as an industry norm.
  6. W3C. Web Content Accessibility Guidelines (WCAG) 2.2, Success Criterion 1.1.1 Non-text Content. W3C Recommendation, 12 December 2024. read 15 October 2026. View on W3CThe normative requirement for a text alternative. Note the criterion requires an equivalent text alternative; the short-plus-long description pairing is a technique rather than part of the criterion.
  7. Springer Nature. Accessibility requirements for figures, images and supplementary materials. read 15 October 2026. View on Springer NatureA publisher's own alt-text expectation. Note the warning that it does not make post-publication corrections to alt text, which makes this a submission-stage task.
  8. Taylor & Francis. An author’s guide to writing good alt text. Author Services. read 15 October 2026. View on Taylor & Francis Author ServicesThe only concrete alt-text length guidance we could verify anywhere.
  9. Ibrahim AM, Lillemoe KD, Klingensmith ME, Dimick JB. Visual abstracts to disseminate research on social media: a prospective, case-control crossover study. Annals of Surgery. 2017. doi:10.1097/SLA.0000000000002277 read 15 October 2026. View on a society-hosted copyThe study usually cited for the effect of visual abstracts: 44 articles, each tweeted twice. Read from a pre-pagination copy on a society website, so volume and page numbers are not given here.
  10. Oska S, Lerma E, Topf J. A picture is worth a thousand views: a triple crossover trial of visual abstracts to examine their impact on research dissemination. Journal of Medical Internet Research. 2020;22(12):e22327. doi:10.2196/22327 read 15 October 2026. View on JMIRThe independent replication, read on the publisher's own site. It found a much smaller effect than the original and, importantly, no significant difference in link clicks.
  11. Buljan I, Malicki M, Wager E, et al. No difference in knowledge obtained from infographic or plain language summary of a Cochrane systematic review: three randomized controlled trials. Journal of Clinical Epidemiology. 2018;97:86–94. doi:10.1016/j.jclinepi.2017.12.003 read 15 October 2026. View on a university repository recordThree randomised trials finding no knowledge gain from an infographic over plain text, though readers preferred it. Read as the repository abstract only.

Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.

Related: types of scientific communication · structuring tables and figures · writing for a high-impact journal.

How Pubrica can help

Pubrica’s scientific illustrators prepare graphical abstracts to the target journal’s specification, including dimensions, aspect ratio, file formats, typography and panel rules, with permissions cleared for every element and alt text drafted for submission. Reformatting a graphical abstract returned by a journal is a discrete, defined task.

Get a free quote

Regulations

Check your project against this rule

  1. 1Your work
  2. 2Your details
  3. 3Send
What do you need?

Material you send is held under signed confidentiality, with access limited to the assigned team and encrypted transfer. It is not used to train anything and is not passed to third-party tools. Ask for the agreement before you send anything you would rather cover first.