The way we write scientific papers doesn’t change very quickly. Once we settled into the AIMRaD (Abstract, Introduction, Methods, Results, and Discussion) format, we pretty much stuck with it: the vast majority of scientific papers use that format, or one of a few minor variations on it.* But we do tinker; and so, there’s a relatively new wrinkle on the A part of AIMRaD. It’s the “graphical abstract”: an illustration that’s supposed to – well, what’s it supposed to do exactly?
To be honest, I really don’t understand the point of a graphical abstract. Here’s what one journal (Ecological Modeling) says it’s for:
I have never, ever – not even once – seen a graphical abstract that usefully “summarized the contents of an article”. I’ll go bigger: I’ve never seen a graphical abstract that I could even understand without also reading at least the text Abstract, if not more of the paper. I’m not sure a graphical abstract that “summarizes the content of the article” is even possible. We invented written language, many thousands of years ago, for a reason: drawing pictures (alone) just isn’t an effective way to communicate complex information.** (There’s some evidence here that this claim has more support than just my interior vibes.) Pictures are also a terrible way to support users searching for articles (which is another major function of an abstract). A graphical abstract might well “support readers in digesting [the] research”; but if that’s its role, then it doesn’t have much to do with the Abstract. Instead, it would be a conceptual illustration that belongs in the Introduction or maybe in the Discussion. That’s nothing new, and calling it a “graphical abstract” just seems to muddy the water.
There’s another possibility in the Ecological Modeling guidance. Could a graphical abstract “draw more attention to your article”? It sure could, both for good and for bad.
For good: there’s some evidence that social-media attention to a paper can increase its impact; and being able to attach a graphical abstract to your post might help with that. But I say “might” for a couple of reasons. First, even the “increase impact” claim is only moderately well supported. And second: the hypothesis that a graphical abstract, in particular, can have such an effect hasn’t been tested; and in many cases I’m willing to bet that a photograph of your study species, site, equipment, etc. would do as well or better. Especially if you happen to work on something charismatic, like giant pandas, or leeches.
But also: attention for bad. That monstrosity above this post? It’s gotten some attention for the paper it, um, “illustrates”; but it’s not the kind of attention you’d want. It’s AI-generated trash (obviously) – illegible text, “carbon sonks”, “concologiitc serves callues”, and more. Look, folks: I think there are useful and appropriate ways to use generative AI in scientific writing. But then there’s this. For the love of whatever deity might be most appropriate, do not do this.
And yet, I halfway understand and even halfway admire the decision the authors of the monstrosity in question made: to outsource their graphical abstract to genAI. If a task is a waste of time, such that even doing the task well won’t accomplish much, and yet you’re pushed to do it anyway – isn’t it perhaps entirely rational to let genAI do a crappy job of it, and move on with your life? And if I had my career over again, I can think of more than a few documents – strategic plans, grant reports, and the like – that I might have been better off treating like the graphical abstract I’m poking fun at. Although I probably would have edited out the carbon sonks.
OK, griping is fun. But I usually try to leaven my griping with something useful. So here goes: I can think of one way that a graphical abstract could be useful – but to a writer, not a reader. I bet every single graphical abstract ever published has been made last in the paper-writing sequence.*** What if, instead, you made it first? What if challenging yourself to illustrate your paper’s logic and results and importance in a single graphic (and you can call it a “graphical abstract” if you like) was a way of helping you find your paper’s story? What if it was just a slightly more formalized version of the office white-board doodles we often fiddle with before we sit down to write? What if we then used it as a guide while writing text, to help make sure the manuscript we produce has the structure, and tells the story, that we decided it should? And when what if, like that office white-board doodle, we then erased the “graphical abstract” rather than inflicting it on our readers? (Whoops, I guess I’m back to griping again. Sorry.)
Now, strangely, nobody has (yet) seen fit to make me the benevolent dictator of scientific publishing; and so my dislike for publishing graphical abstracts doesn’t take them off the table. Many journals now encourage graphical abstracts; a few even require them. So if you need to make one (or if you disagree with my take, and thus actually want to make one), there’s some advice available on how to make good ones. Well, better ones, anyway. You can start with Lee and Yoo 2023 and Jambor and Bornhäuser 2024. And if you’re making a graphical abstract, you might as well have fun with it; it seems like a place you could sneak in a little humour or beauty. Just, please, no carbon sonks and definitely no concologiitc serves callues. Thanks.
© Stephen Heard Sept. 23, 2025
The less sassy parts of this post are based in part on material from the forthcoming 3rd edition of The Scientist’s Guide to Writing. I’m afraid it’s not imminently forthcoming; expect it some time in mid- to late 2026.
Image: © 2025 Elsevier B.V. from Zhao et al. 2025, fair use (for criticism).
*^Reviews aren’t typically AIMRaD; nor are notes and comments. But among the regular primary-results papers that are our scientific bread and butter, it’s hard to find something that isn’t fundamentally AIMRaD. The methods-last paper is probably the most common apparent exception, but if you read one carefully, it’s hardly an exception at all.
**^I’m looking at you, Ikea assembly instructions.
***^Which isn’t a bad thing, if you’re thinking of one as an actual abstract. I write my Abstracts last, on the grounds that an Abstract should summarize a paper, and it’s really hard to summarize something you haven’t written yet. But that logic disappears if the “graphical abstract” isn’t actually an abstract.