Nikon re-reviews its microscopy video winner over AI use

Nikon is re-reviewing the video that won its 2026 Small World in Motion competition after scientists raised questions about how much AI went into it. The winning entry shows the abnormal beating of airway cilia in a child with primary ciliary dyskinesia, a rare genetic disorder, and took first place in the microscopy contest in September.

The criticism started when the competition posted the video on LinkedIn. Edward Phelps, an associate professor at the University of Florida, wrote that there were "many serious problems with this video", saying some of the biological structures do not behave as they should, and that the microscopy community deserved an explanation from Nikon. Ian Donovan, a doctoral student at UT Southwestern Medical Center, flagged an embedded SynthID watermark, the marker Google uses to label synthetic content, CNN reported.

The entrant, Ning Xu, replied on the same post. At the time of the competition he was a researcher at Tsinghua University in China; he has since joined the National University of Singapore as a research fellow, according to Nikon's own profile of Xu, which CNN also cites. Xu said AI was not used to generate the experimental movie, the cilia, or their motion. "It was applied afterwards to the reconstructed grayscale data to distinguish and colour structures with similar morphology," he wrote. He added that he is preparing a fuller description of the optical system and a quantitative analysis of the ciliary motion for publication.

Nikon said it is "carefully re-reviewing the information" Xu provided, and that he is "respectfully complying" and has submitted detailed documentation covering the microscopy equipment, the imaging methods and the processing techniques used to create the source video.

The dispute is not really about whether generative tools touched the file. It is about where the line sits between processing and creation in scientific imaging. Reconstructing a grayscale image and colouring structures to make them legible is standard practice in microscopy. Generating detail that was not captured is a different act, and it is the one that would put a competition entry in conflict with its own rules. Xu's account places his work on the first side of that line; the reviewers' questions about structures behaving unexpectedly place it in doubt.

The watermark is the awkward part. SynthID is applied to output from Google's generative models, so its presence in an entry that was not generated by a Google model either points to a processing step Xu has not described in that much detail, or to provenance being harder to establish than the competition's rules assume. Nikon has not said what it has concluded, and has not announced a change to the result.

Nikon has not said when its review will conclude; the entry stands until the organisers say otherwise.