New research integrity AI tool added to Springer Nature’s growing portfolio A new AI tool to identify irrelevant references in submitted manuscripts has been launched for use across Springer Nature’s journals and books.

A new AI tool to identify irrelevant references in submitted manuscripts has been launched for use across Springer Nature's journals and books. 


The tool, announced on 7 April 2025, is the latest AI-driven tool that has been developed in-house at Springer Nature to weed out problematic submissions and ensure the veracity of the publication record. 

The tool will be used by Springer Nature’s Research Integrity Group (RIG) to assess submissions to almost all journals and books published by Springer Nature, analysing the relevance of each reference used. If a number of references are identified as irrelevant, the submission will be flagged to RIG who will manually check the manuscript and decide whether the submission should be withdrawn.

The irrelevant reference checker tool has undergone multiple rounds of testing and validation to ensure it provides a precise and reliable assessment of references across academic disciplines, and training and development of the tool will be ongoing.  Human oversight will always remain in place, in line with Springer Nature’s AI Principles

The launch of the irrelevant reference checker tool comes just ten months after the launch of two other AI tools to detect problematic submissions that Springer Nature has developed in-house.  The first, Geppetto, supports the identification of papers that contain AI-generated fake content, the second, SnappShot, those that contain problematic images.

On 30 April, 2025 the company announced it would donate Geppetto to STM, where it will be integrated into the STM Integrity Hub, an industry-wide initiative that supports publishers in ensuring the integrity of their published content, as part of its mission to develop and trial tools that publishers large and small can use to screen submissions for indicators of compromised content.