Should you use AI animation for your research?

If you're weighing up whether an AI-generated animation would work for your project, the honest answer is that it depends on what the video is for. For a quick social media clip that disappears after a few days, maybe it matters less. But if the video carries your name, your institution, or a finding you want people to believe, AI animation usually costs you more than it saves. The real cost isn't money. It's trust.

The argument for AI is usually simple: faster, cheaper, good enough. For research communication, "good enough" isn't good enough.

People are getting better at recognising AI

People are becoming increasingly familiar with the patterns of AI-generated content. The slightly plastic surfaces, the flat lighting, that faint uncanny quality where nothing is obviously wrong and yet something feels different. Most people won't know exactly why it looks odd. They'll just know it feels like it came from a system rather than from someone making deliberate creative choices. And once someone suspects something was made by AI, it can change how they interpret the rest of what they're seeing: if this part was done the easy way, what else was?

For many commercial products, that might be a minor issue. For research, that's a much bigger problem. Your work only matters if people take it seriously, and a video that signals corner-cutting makes them wonder about the rigour behind it. You probably spent years getting the research right. It seems a shame to undermine that with the thing that's supposed to explain it.

Everything ends up looking the same

There's a simpler problem too. AI tools drift towards one house style. Put in a similar prompt and you get a similar look back: the same palette, the same rendering, the same mood thousands of other people are generating around the same time.

Every paper is different. The animation should feel like it was made for that paper, not pulled from the same template as hundreds of others.

For research, that lack of specificity is not just an aesthetic problem. It becomes a credibility problem.

We've already seen the consequences

AI visuals have already blown up inside serious journals.

In February 2024, the journal Frontiers in Cell and Developmental Biology retracted a paper after readers spotted that its figures had been made with Midjourney. One image showed a rat with anatomically impossible parts and labels that were complete gibberish: "iollotte sserotgomar," "testtomcels," "senctolic." Concerns about the figures were raised during the publication process, but the paper was published before later being retracted.

And it wasn't a one-off. In April 2026, the New England Journal of Medicine retracted a clinical image after the authors admitted using an AI tool to alter the photo, its first retraction since 2020. The alteration was small, but undisclosed AI manipulation was enough for the authors to retract the report.

The same question applies when the video carries your paper, your grant, or your institution's name. Retractions make headlines. More often, the problem is quieter: a video that looks convincing while subtly getting things wrong, that nobody flags, that quietly misleads the exact people you were trying to reach.

Getting it right is the whole job

Explaining research isn't decoration. Every visual is a claim about what the study says. When I animate a paper, the hard part isn't making it look nice, it's making sure the picture says exactly what the data says and not a fraction more.

Take a recent animation I made for IARC (the WHO's cancer agency) about a lung-cancer screening study. The authors sat down and reviewed the storyboard with me, and they caught small things that mattered scientifically, points a viewer could easily have misread, and we reworked each one until the animation matched the paper precisely. That doesn't come out of a prompt box. It comes from a person reading the actual study, understanding what it claims, and going back and forth with the researchers about what's true and what isn't.

The environmental and ethical cost

If your work touches sustainability, public health, or ethics, reaching for generative AI to explain it isn't a neutral choice, and your audience will feel the contradiction.

The environmental cost of generative AI is real, and animation is where it piles up. A minute of video is well over a thousand frames, and getting a usable result means generating far more than that once you count the discarded takes and regenerations. Stretch that across a whole animated video and it adds up fast. According to MIT, data-centre electricity use is expected to more than double between 2022 and 2026, and many regions still rely heavily on fossil fuels for electricity generation. These facilities also use vast amounts of water for cooling: Google reported using 5.6 billion gallons in 2022, up 20% in a single year, and Microsoft reported a 34% rise, as their data-centre operations expanded.

There's also the question of whose work it's built on. Generative AI systems were trained on enormous datasets containing creative works, and many artists have raised concerns that their work was used without individual permission or compensation. When you commission AI-generated output, that is part of the wider system you are supporting. For a lot of the researchers I work with, that alone is reason enough to keep their name away from it.

The judgement is the point

Research animation is not simply the process of generating images. The difficult part is deciding what the viewer should actually understand from those images. Every choice involves interpretation: what to simplify, what to emphasise, and what could accidentally create a misunderstanding. That judgement comes from understanding the research itself.

What actually protects your work

A research explainer exists to make something complicated understandable while staying faithful to the research, for people who need to trust it. That happens when a real person reads your paper, asks the awkward questions, gets the science right, and gives the work a look that's yours and nobody else's.

Yes, that's slower than typing a prompt, and it should be. That extra time isn't wasted. It's what you're actually paying for. When the goal is to make people understand and trust your research, the process behind the animation matters as much as the final image.