Graphical Abstract Requirements (2026): Size, Format and What Each Publisher Actually Specifies

Four major publishers publish a graphical abstract specification, and they specify four different shapes. Elsevier wants a wide banner at two and a half times as wide as it is tall. Cell Press wants a perfect square. ACS wants something that fits in a box a little over three inches across. MDPI states its minimum height before its width, which is a good way to submit an image rotated ninety degrees from what you meant.

None of those four shapes is the shape of a slide, which is what most graphical abstracts are built on. That is the single most common reason one comes back from production.

This page gives each publisher’s numbers with the document they came from, and says plainly where a publisher states nothing rather than filling the gap with a number that sounds right.

Graphical abstract requirements at a glance

Every row below is the publisher’s own number, from the document named in the last column. Where a cell says not stated, the document genuinely does not state it.

What is a graphical abstract?

A single image that summarises the take-home message of a paper, displayed beside or beneath the text abstract on the journal’s website. Elsevier’s author page describes it as a concise pictorial summary designed to appeal to an interdisciplinary audience, shown in online search result lists, the online contents list and the article page on ScienceDirect. Both Elsevier and MDPI note the same practical consequence: it typically does not appear in the article PDF or in print. It is a web asset.

That changes how you should design it. The thing being optimised is a thumbnail in a results list, not a page in a printed issue. Elsevier is explicit that its image gets scaled proportionally into a 500 by 200 pixel rectangle on ScienceDirect, which is why it asks for fonts large enough to survive reduction to a 200-pixel-high window.

Two names describe the same object. Elsevier uses graphical abstract and visual abstract interchangeably on the same page. ACS calls it a Table of Contents graphic or abstract graphic, and its author guidelines use both terms. Cell Press abbreviates it to GA throughout its guidelines. If you are searching a journal’s instructions and find nothing under one term, search the others before concluding the journal does not want one.

If you want to see what the format looks like across fields before reading the specifications, our graphical abstract examples collection shows the layouts that recur.

What size does Elsevier require?

A minimum of 1328 × 531 pixels, width by height, at a minimum resolution of 300 dpi. If you submit larger, Elsevier asks you to keep the same ratio, which it expresses as 500 wide by 200 high. That is 2.5:1, and it is unusually wide.

The rest of the Elsevier specification, from the same page:

• Fonts must be Times, Arial, Courier or Symbol, at a size large enough to remain legible once the image is reduced to fit a 200-pixel-high window.
• Preferred file types are TIFF, EPS, PDF or MS Office files.
• The image should read from top to bottom or left to right, with a clear start and end.
• No additional text, outline or synopsis. Any text or label has to be part of the image file itself.
• No heading reading “graphical abstract” inside the image, and no unnecessary white space.
• Submit it as a separate file in the online submission system.

Elsevier also directs you to the individual journal’s guide for authors for specific policies, and that instruction is load-bearing. Elsevier publishes thousands of titles under this one page, and whether a graphical abstract is required, optional or unavailable is decided per journal, not by this document.

Cell Press wants 1200 pixels square, and usually requires one

Cell Press is part of Elsevier but keeps its own graphical abstract guidelines, and the numbers differ. The submitted image should be 1200 pixels square at 300 dpi, with Arial at 8 to 12 points, in TIFF, PDF or JPG. Smaller fonts, the guidelines say, will not be legible online.

The requirement that catches people out is not a number. The Cell final submission page states that a graphical abstract is optional for Cell Systems, Current Biology, Patterns, Neuron and Newton, and required for all other journals in the family. For Cell, Molecular Cell, Immunity, Cell Reports, Developmental Cell and the rest, it is part of the submission, not a marketing extra you can skip when the deadline gets close.

The content rules are stricter than any other publisher’s. A Cell Press graphical abstract must:

• Be distinct from any model figure or diagram in the paper itself.
• Contain no data items of any type. All content has to be in graphical form, which rules out pasting in a plot.
• Emphasise the new findings from this paper without carrying excess detail from previous literature.
• Give a visual indication of biological context, such as subcellular location, tissue or cell type, or species.
• Highlight one process, or make one point clear.
• Contain only elements that are original or licensed for commercial use.

Cell Press also accepts graphical abstracts made in BioRender, provided the author follows its template and holds the correct permissions. That is a rare piece of explicit tool guidance in an author instruction, and it matters for the AI question further down.

What ACS specifies for a TOC graphic

ACS states the size in inches rather than pixels, because the graphic is laid out at a physical size. Submit it at the actual size to be used, fitting within an area no larger than 3.25 by 1.75 inches, which the guidelines give as approximately 8.25 by 4.45 cm. Type should be sans serif such as Helvetica, preferably 8 pt and never smaller than 6 pt. The file should be TIF at 300 dpi for colour and 1200 dpi for black and white.

Those two numbers together fix the pixel dimensions, though ACS does not state them as pixels: 3.25 inches at 300 dpi is 975 pixels wide, and 1.75 inches is 525 pixels high. Treat that as arithmetic from the ACS specification rather than as something ACS publishes.

The ACS content rules are the most restrictive of the four, and several are absolute:

• The graphic must be entirely original, unpublished artwork created by one of the coauthors.
• It must not include a photograph, drawing or caricature of any person, living or deceased.
• No postage stamps, no currency from any country, no trademarked items, including company logos, images and products.
• Text should usually be limited to labelling compounds, reaction arrows and diagrams. Long phrases or sentences should be avoided.
• Avoid a graphic that already appears within the manuscript.

On whether it is required, ACS is unambiguous where Elsevier defers to the journal. The guidelines open by stating that a graphic must be included with each manuscript, for display in the table of contents and the abstract. The JACS author guidelines point to the same document for specifications. The version we checked carries the date 28 February 2024 on its first page.

MDPI states its size height-first

The minimum required size for an MDPI graphical abstract is 560 × 1100 pixels, and the Layout Style Guide writes that as height × width. So it is 1100 wide and 560 tall, roughly 1.96:1. Read that ordering carefully, because it is the opposite of the convention Elsevier uses on its page three paragraphs earlier in your browser history.

MDPI accepts PNG, JPEG or TIFF, and recommends Times, Arial, Courier, Helvetica, Ubuntu or Calibri. It states no DPI at all, which is worth saying out loud because most pages that claim to summarise MDPI’s requirements supply one anyway. If you want a number to work to, the pixel minimum is the binding constraint, and exporting at 300 dpi satisfies every other publisher you might submit to next.

MDPI’s originality rules are phrased more precisely than most, and they close three specific loopholes. A graphical abstract should not be exactly the same as any figure in the body of the paper, nor a simple superposition of several subfigures, nor a combination of the text abstract and a picture. It must be original, unpublished artwork, with no postage stamps, currency or trademarked items, and no copyrighted material. As with Elsevier, do not put a “Graphical Abstract” heading inside the image.

Does Nature accept a graphical abstract?

The Nature formatting guide does not mention them. It is a long set of author instructions that refers to figures dozens of times, without once using the term graphical abstract or visual abstract. The Nature final-submission page does not use either term. Neither does the Nature Communications submission guide.

An absence is weaker evidence than a statement, so here is a statement from the same publisher: Scientific Reports, a Springer Nature title, says outright in its submission guidelines that it does not support graphical abstracts. We cover that alongside its other specifications in our Scientific Reports figure requirements guide.

The practical reading is that a graphical abstract is out of scope for a Nature submission unless the specific journal asks for one in its own instructions. Springer Nature is a large publisher and some of its non-Nature journals do request them, so check the title rather than generalising from the brand. What Nature does specify, in detail, is the artwork itself — lettering size, column width, line weight and file format — which we go through in our Nature figure and impact factor guide.

If you have already made a graphical abstract for another journal and are now submitting to a Nature title, there is nowhere to upload it. The effort is better spent on the first figure, which is the one an editor sees.

Wiley, and the limits of a publisher-level answer

We could not verify a Wiley-wide graphical abstract specification, and we are not going to invent one.

Wiley Author Services publishes a page at the address authors.wiley.com/author-resources/Journal-Authors/Prepare/graphical-toc-abstract.html, which is the canonical location cited across the web for Wiley’s graphical table of contents requirements. On 13 August 2026 that page returned an HTTP 200 with an empty document body, both to a plain fetch and to a full browser session that renders JavaScript. The Angewandte Chemie author guidelines page loaded, but served only navigation.

What that means for an author is straightforward. Wiley’s graphical table of contents specifications are set at journal level, and the journal’s own author guidelines are the document that binds you. Any page quoting a single Wiley pixel size is quoting something it did not verify, or quoting one journal and presenting it as the publisher’s rule.

Can you use AI to make a graphical abstract?

Elsevier answers this directly, and its answer is narrower than most authors assume. Its policy on generative AI in figures, images and artwork sorts images into three categories, and graphical abstracts sit in their own category alongside cover art, on the grounds that they summarise the article rather than form part of its scientific content.

For that category the policy states that general-purpose generative AI image tools must not be used to create graphical abstracts. It goes on to encourage dedicated scientific illustration tools instead, naming BioRender and Mind the Graph as examples, on the reasoning that such tools carry well-defined licensing terms for the icons and elements they supply. Where an AI tool is used, the policy asks for the tool to be named in the image caption. Cell Press applies the same Elsevier policy and repeats the rule in its own guidelines, while explicitly accepting BioRender output that follows its template.

The line being drawn is not AI versus no AI. It is between a text-to-image model that invents pixels with no traceable rights, and an illustration tool with curated assets and clear licensing. If you are choosing a tool with that distinction in mind, our comparison of BioRender alternatives covers which ones produce licensed, editable vector output and which do not.

Two things follow for anyone using any tool at all. Keep a record of what made each element, because the caption disclosure requires it. And check the licence on the free tier of whatever you use, because Elsevier notes that free versions of illustration tools are sometimes limited to educational or internal use and may not permit publication.

Does a graphical abstract actually increase visibility?

The honest answer is that the evidence is thinner than the marketing.

Elsevier states on its author page that research has shown articles with graphical abstracts see increased views and social media activity, and that the average annual use of an article is doubled compared with those without a visual abstract.

An independent study reached the opposite conclusion. Pferschy-Wenzig and colleagues, publishing in Molecules in 2016, compared papers published in that journal with and without a graphical abstract between March 2014 and March 2015, across several output measures. Papers published without a graphical abstract performed significantly better on PDF downloads, abstract views and total citations. The authors described it as, to their knowledge, the first empirical study of whether graphical abstracts attract attention to publications, and noted that the claimed effectiveness is asserted frequently by publishers and blogs without a thorough test. The paper carries a 2016 erratum, which corrects a percent sign in one table column and leaves the finding unchanged.

One study in one journal does not settle the question, and a journal’s readership and subject mix plausibly matter. But it is enough to justify a straightforward position: make a graphical abstract because the journal requires one, or because you want a self-contained summary you control for talks and posters. Do not make one on the promise of doubled usage.

How to build one that survives production

The four specifications disagree on almost every number, but the rules they share are the ones that actually cause rejections.

Compose at the target ratio from the start. A square Cell Press abstract and a 2.5:1 Elsevier banner are not the same artwork at different scales. Cropping one into the other leaves either dead space or amputated labels. Decide the journal first.

Set the type at the size it will be viewed. Cell Press gives 8 to 12 pt in Arial. ACS gives 8 pt preferred and 6 pt absolute minimum. Elsevier gives no point size but tells you the display window is 200 pixels high, which is the same instruction in different units. Zoom your file to the real display size and read it on screen before submitting.

Make it new artwork. All four publishers forbid reusing a paper figure, in slightly different words. Cell Press goes furthest by banning data items entirely, so a graphical abstract with a bar chart in it fails.

Clear the rights on every element. ACS requires the whole graphic to be original work by a coauthor. MDPI and ACS both name postage stamps, currency and trademarked items, which is a specific enough list to suggest they see them regularly. Icon sets from an unlicensed source are the usual failure.

Leave the label off. No “Graphical Abstract” heading inside the image, per both Elsevier and MDPI. The journal adds that itself.

Export TIFF if you are unsure. It is the only format all four publishers on this page accept.

Keep the source file editable. The reason a resubmission to a different journal is expensive is almost never the artwork. It is that the only surviving copy is a flattened PNG, so changing a 2.5:1 banner into a 1200 pixel square means redrawing it.

Frequently asked questions

What size should a graphical abstract be? There is no single size. Elsevier asks for at least 1328 × 531 px at 300 dpi in a 2.5:1 ratio, Cell Press for 1200 px square at 300 dpi, ACS for a graphic fitting within 3.25 × 1.75 in at display size, and MDPI for at least 560 × 1100 px stated height first. Build for the journal you are submitting to.

Does Cell require a graphical abstract? For most Cell Press titles, yes. It is optional for Cell Systems, Current Biology, Patterns, Neuron and Newton, and required for all other journals in the family.

Does Nature accept a graphical abstract? The Nature formatting guide never mentions one, nor does the Nature final submission page or the Nature Communications submission guide. Scientific Reports, in the same publishing group, states plainly that it does not support them.

Can I use an AI image generator for a graphical abstract? Not a general-purpose one, under Elsevier’s policy, which directs authors to dedicated scientific illustration tools and asks that the tool be named in the caption. Cell Press accepts BioRender output that follows its template.

Can I reuse Figure 1 as my graphical abstract? No. Cell Press requires it to be distinct from any model figure in the paper and to contain no data items at all, MDPI rules out both an identical figure and a superposition of subfigures, and ACS advises against a graphic already in the manuscript.

Does it appear in the PDF? Usually not. Elsevier and MDPI both state that the graphical abstract is shown on the online article and typically absent from the print version and the PDF.

What resolution does MDPI require? None is stated. The Layout Style Guide gives a pixel minimum and file formats but no DPI. Exporting at 300 dpi keeps the file usable elsewhere.

Which file format works everywhere? TIFF. Elsevier, Cell Press, ACS and MDPI all accept it, though ACS wants 1200 dpi for black and white work.

Where to go next

The specifications on this page are the part you can check in an afternoon. The part that costs time is the artwork itself, and specifically whether it was built in a form you can revise when the paper moves to a different journal with a different ratio.

If you are working out which tool to build it in, our comparison of figure tools is honest about which ones export licensed, editable vector output. If the same figures also have to work at wall size, our research poster guide covers the identical legibility problem at the opposite scale. And if the journal you are targeting also specifies its ordinary figures tightly, the two worked examples on this site are our PNAS figure and impact factor guide and our Scientific Reports guide, which between them cover a journal that specifies almost everything and one that specifies almost nothing.