Springer Nature
Geppetto and SnappShot AI screening of manuscripts
Springer Nature said its Geppetto tool caught hundreds of fake papers before publication
Problem
Paper mills sell fake research, and AI-generated text and manipulated images make such papers harder to spot in peer review.
Approach
The publisher built two tools in house. Geppetto splits a paper into sections and scores each for the likelihood of AI-generated text. SnappShot checks PDFs for duplicated gel and blot images. High scores go to staff for manual review, and neither tool decides on its own. After pilots, both were being rolled out in 2024, with SnappShot planned to cover other image types.
Results
Takeaway
The tools only flag papers, so editors still make every decision. The count of fake papers comes from the publisher and was not independently checked.
Categories
- Size
- Large enterprise
- Industry
- Media and entertainment
- Business function
- Production and qualityInternal work and knowledge management
- Innovation type
- Process innovation
- AI technology
- Predictive analytics and MLComputer vision
- Adoption
- Built in-house
- Goal
- Quality and risk
- Evidence
- Company's own claim
Sources
Source summaries
The press release announces the rollout of Geppetto and SnappShot after successful pilots. It says Geppetto already identified hundreds of fake papers and that both tools trigger human checks rather than decide.
The article describes both tools and repeats the publisher's claim of hundreds of fake papers flagged. It notes the publisher did not explain how Geppetto works.
Full details are in the original sources.