Written by 16 November 2026
We're thrilled to share that Imagetwin has won the 2026 ALPSP Award for Innovation in Publishing, announced at the ALPSP Annual Conference in Manchester. It's a huge moment for our team, and a strong signal that image integrity is finally getting the attention it deserves in scholarly publishing.
Tell us about your organization
Imagetwin is building an AI-powered image integrity software. We help publishers, universities and research institutions catch duplicated, manipulated, plagiarized and AI-generated images in scientific manuscripts, before those images make it into the published record.
The company grew out of research at TU Wien, where co-founder Markus Zlabinger began exploring how to detect image plagiarism in scientific publications during his own academic work. After encouraging early feedback from research integrity experts, including Elisabeth Bik, and a place in TU Wien's incubation programme, that research became Imagetwin, which officially launched in January 2022.
Today Imagetwin is used by individuals and teams at more than 1,500 academic organizations worldwide, including 9 of the top 10 academic publishers, and we're proud members of STM and COPE.
What is the project/product you submitted for the Award?
We submitted Imagetwin itself, our core screening platform. It scans the figures in a manuscript and checks them against a database of more than 160 million academic images to flag duplication, manipulation, plagiarism and AI-generated content, all within seconds. Publishers can plug it directly into their existing editorial workflow (we integrate with ScholarOne, Editorial Manager, Morressier, Cactus and others), so screening becomes a routine, automated step rather than extra manual work for already-stretched editorial teams.
Tell us a little about how it works, and the team behind it
Imagetwin runs on computer vision and machine learning models trained specifically for scientific figures: microscopy, gel bands, light photography, FACS plots and more. Each image type behaves differently, so instead of one generic model, we've built and tuned separate detection engines per type, which is a large part of why detection accuracy is the strongest in the industry, up to 99%, depending on the image.
The team behind that work brings together expertise in computer vision, machine learning and research integrity, led by Markus as CTO on the technical side and myself on the business and partnerships side. We've stayed deliberately close to the community we serve throughout, treating publisher and researcher feedback as an ongoing input into how we build, not an afterthought.
In what ways do you think it demonstrates innovation?
Text plagiarism checks became routine in publishing years ago, but image integrity never got the same treatment, largely because figures are so much harder to standardize and screen at scale than text. Imagetwin was built specifically to close that gap, and we've gone deep rather than wide: rather than a long list of shallow checks, we've focused on best-in-class detection for the hardest image types first.
The results back that up. In an independent review of 715 papers, 16% contained image integrity issues, and Imagetwin caught 87% of them, compared with 55% caught by a human expert working alone. A 2025 pilot with the American Society for Microbiology screened 2,627 manuscripts across 15 journals and flagged concerns in 15.6% of them, catching issues pre-publication in around 1.5 hours of staff time versus roughly 10 hours to resolve the same kind of issue after the fact. Radboud University Medical Center used Imagetwin to retrospectively screen 608 papers from its neuroscience archive and found image-related concerns in roughly 40% of them, an audit that would have been effectively impossible to do manually at that scale.
What are your plans for the future?
Image integrity is the foundation we've already built and are constantly improving, adding new image types and adapting our detection as new AI generation models emerge. From now on, we're also expanding beyond images into a more comprehensive platform that will cover different integrity checks and allow deeper integrations with manuscript submission systems and expert-review services. The goal hasn't changed since we started: give publishers and institutions a way to catch problems before they're published, rather than paying to clean them up, at far greater cost, afterwards.
Want to see Imagetwin in action, or talk through what image screening could look like at your organization? Visit imagetwin.ai or get in touch.
linkedin.com/company/imagetwin · x.com/imagetwinai
About the author: Patrick Starke is CEO and co-founder of Imagetwin, the AI-powered image integrity platform used by 9 of the world's 10 leading academic publishers. Patrick combines economics and deep-tech startup experience to advance Imagetwin's research integrity mission.