To make a graphical abstract, check your journal’s size and AI rules, pick the one finding readers should remember, sketch a layout that reads left to right, draw it in a tool your journal accepts, then test it on someone who has not read the paper.
A graphical abstract is a single image that sums up a paper’s main message. Elsevier shows it in search results, contents lists and the article page on ScienceDirect, but usually not in the PDF, and it cites research showing articles with a graphical abstract get double the average annual use. Many graduate students, including those on Western New York campuses, first meet the request with their debut manuscript.
To keep the steps concrete, this guide runs one real study through all five of them: “Weak anthropogenic electric fields affect honeybee foraging” by Victoria Mallinson, Fraser Woodburn and Liam O’Reilly (iScience, 2025), an open-access CC BY 4.0 paper published with its own graphical abstract. In field experiments in urban meadows, the team exposed flowers to weak electric fields of the kind people create and counted how often honeybees landed. They also measured the fields near high-voltage power lines. The study is only a worked example here; its findings belong to the authors.
Step 1: Check Your Journal’s Size, File and AI Rules
Start with your journal’s guide for authors, because graphical abstract size, file type and AI rules differ by publisher. Three large publishers ask for the following in their own author guidance, checked in October 2026:
| Publisher | Size | Resolution and files | Text | Source |
| Elsevier | At least 1328 × 531 px (width × height); larger images keep the same 5:2 ratio | 300 dpi minimum; TIFF, EPS, PDF or MS Office files | Times, Arial, Courier or Symbol, big enough to read at 200 px high | Elsevier graphical abstract guide (current page) |
| Cell Press | 1200 px square | 300 dpi | Check the journal | Cell Press CrossTalk blog, 2018 |
| ACS (called a TOC graphic) | No larger than 3.25 × 1.75 in (8.25 × 4.45 cm), submitted at actual size | TIF at 300 dpi for color, 1200 dpi for black and white | Sans serif such as Helvetica, 8 pt preferred, never under 6 pt | ACS TOC/Abstract Graphics guidelines, updated February 2024 |

Figure 1. Canvas shapes for three publishers, drawn to scale from each one’s author guidance (checked October 2026).
Read the AI policy in the same sitting. Elsevier’s generative AI policy, updated June 2026, says: “General-purpose generative AI image tools must not be used to create graphical abstracts.” It points authors to dedicated scientific illustration tools and asks that any AI tool, its version and how it was used be disclosed in the image caption. ACS asks for original artwork by a coauthor, with no pictures of people and no company logos, and suggests not reusing a figure that already appears in the manuscript.
Worked example. iScience is a Cell Press journal, Cell Press asks for a square, and the bee study’s published graphical abstract is square (Figure 2). Had the same paper gone to a journal asking for Elsevier’s 5:2 strip, the two scenes would have had to sit in a wide row instead.
Step 2: Pick the One Finding You Want Remembered
A graphical abstract carries one take-home message, so write that message as a single sentence before you open any software. “Ten Simple Rules for Designing Graphical Abstracts” (Jambor and Bornhäuser, PLOS Computational Biology, 2024) puts the key message first for this reason.
Then list the three to five things a reader must see in the graphical abstract to believe it: usually the starting point, what you did, and what changed. Everything else stays in the paper.
Check how much detail your journal wants. ACS describes the graphic as a quick impression of the paper “without providing specific results,” so a headline number may not suit every journal.
Worked example. The authors’ first highlight already states the message: “Weak anthropogenic-like electric fields reduce honeybee floral landing.” Four elements carry it: bees, flowers, an electric field and the power lines that make such fields part of daily life. The exact landing reductions, the AC and DC comparison and the field measurements all stay in the paper, and the published graphical abstract shows no numbers at all.
Step 3: Sketch the Layout and Reading Order
Arrange the elements in the order people read: left to right or top to bottom for a process with a clear start and end, a clockwise circle for a cycle, and two side-by-side panels for a before-and-after comparison.
Sketch it in pencil inside a box at your journal’s shape. Four rules keep the sketch readable:
- Group related items in one box or cluster. The Ten Simple Rules paper cites the “magical number 7,” the idea that people take in about seven items at once.
- Give each arrow or line style one meaning, and keep it the same everywhere.
- Use one main color and one accent, and check that the figure still works in grayscale and for color-blind readers.
- Take icons from one library so line width, color and level of detail match.
Worked example. The bee study’s graphical abstract uses the side-by-side comparison layout. On a square canvas, two patches of purple flowers sit next to each other, echoing the paper’s setup of two catmint plants with one exposed to a field. At left, bees land on the flowers. At right, an orange wave rises beside an electrode, and dashed flight paths curve away from the patch. A row of power pylons runs across the background. The single bright accent, orange, marks the electric field, and the image carries no text, so the two halves do the explaining.

Figure 2. The published graphical abstract of Mallinson, Woodburn and O’Reilly, “Weak anthropogenic electric fields affect honeybee foraging,” iScience, 2025 (doi:10.1016/j.isci.2025.112550). Reproduced unchanged under CC BY 4.0.
Step 4: Draw It in a Tool Your Journal Accepts
Most graphical abstracts come from one of three kinds of tools, and your journal’s AI policy decides which ones you can use:
- Drawing and slide software. PowerPoint, Inkscape (free) and Adobe Illustrator give you full control and the most manual work. The Ten Simple Rules authors say PowerPoint is enough for most graphical abstracts if you resize the canvas and export a PDF. Elsevier’s graphical abstract page offers a free template and published examples.
- Scientific icon libraries. BioRender and Mind the Graph supply scientific icons and templates, and Elsevier’s policy names both as dedicated illustration tools. Check the license: BioRender’s pricing page lists “No commercial use or publication” on its free plan, and Elsevier warns that free versions of illustration tools may be limited to education or internal use.
- AI drafting tools. These turn a pasted abstract into a first layout. One example is the graphical abstract maker from Vizcept, an AI scientific figure generator. The first figure is free without an account, as a watermarked 2K PNG; watermark-free files, 4K and SVG start at $14.90 a month or a $19.90 credit pack that does not renew.
Worked example. For the bee study, you would paste the paper’s abstract into the tool, as in Figure 3, then check the draft against the paper.

Figure 3. The Vizcept graphical abstract maker with an abstract pasted in (screenshot, October 2026). The abstract is from Mallinson et al., iScience, 2025, CC BY 4.0.
Where AI drafts fall short: Vizcept’s graphical abstracts are drawn by an image model, so a label, arrow or number can come out wrong. Check each one, and confirm your journal allows AI-assisted figures; for an Elsevier journal, use the draft only as a layout plan. ChatGPT’s image generator falls under the same Elsevier rule.
For a data panel or a trial flow, Vizcept’s free violin, box and scatter plot makers draw charts from pasted numbers without a watermark, and its CONSORT flow diagram generator lays out the boxes in code, so each count sits where you typed it.
Step 5: Test, Export and Submit
Before you upload, show the draft graphical abstract to someone who has not read the paper. If they cannot name the main finding in one sentence, cut elements until they can. ACS suggests this test, plus a printout to check that the text is readable.
Next, view the figure at its real display size. Elsevier scales graphical abstracts into a 500 × 200 px window, so text that is sharp full-screen may blur. Then export to your journal’s specification and upload it as a separate file. Elsevier asks for no heading reading “graphical abstract,” no unnecessary white space, and all text inside the image itself.
Finally, if any AI tool touched the figure, disclose it the way your publisher asks, naming the tool, its version and what it did.
Worked example. Show the bee graphical abstract to a colleague outside ecology and ask one question: what does the orange wave do? If the answer is “it keeps bees off the flowers,” the figure works. Then shrink it to thumbnail size and check that the dashed flight paths are still visible.
Frequently Asked Questions
Can ChatGPT make a graphical abstract?
ChatGPT can generate an image, but Elsevier journals bar general-purpose AI image tools for graphical abstracts. Other publishers set their own rules, so check the guide for authors.
Is there a free tool that can create graphical abstracts?
Yes. Inkscape is free and open source. Free plans of icon libraries and AI tools may add watermarks or rule out publication, so read the license first.
What software can I use to create graphical abstracts?
Graphical abstracts are typically made with the same software as other figures: PowerPoint, Inkscape or Adobe Illustrator for drawing, BioRender or Mind the Graph for scientific icons, and AI drafting tools where the journal allows them. Choose by your journal’s AI policy first and budget second.




Leave a Reply