# Springer AI reference checks are not working

> Source: <https://veruscite.com/blog/springer-ai-checks-are-not-working>
> Published: 2026-09-09 16:28:04+00:00

Being a little salty while trying to sell VerusCite, one of the hiccups has been different publishing groups claiming they already have internal tools that do the same work.

    Here, for example, is a
    [Springer Nature press release in April 2025](https://group.springernature.com/gp/group/media/press-releases/new-research-integrity-ai-tool/27769148)
    claiming they have an internal AI tool to check the relevance of citations. While that is
    technically different from determining whether a citation exists at all, someone may need to
    fill me in on how exactly you determine that a hallucinated reference is relevant.

    Just for some evidence that Springer’s internal tools are
    *definitely not working*, here are articles published in August or September of 2026
    where VerusCite
    [identified hallucinated citations](https://veruscite.com/hallucinations)
    (and I confirmed each with a second human check).

### September 2026

**Large language models outperform humans at estimating society’s everyday norms**

    Kimmo Eriksson, Simon Karlsson, Irina Vartanova, & Pontus Strimling

*Communications AI & Computing* ·
    [10.1038/s44488-026-00018-8](https://doi.org/10.1038/s44488-026-00018-8)

[VerusCite report](https://veruscite.com/share/u0hPBigMx87cuYrhNnLZre1yBh-lOoEYfMp1GLHA8_U)

**Generative artificial intelligence produces differential academic outcomes across socioeconomic, linguistic, and disability contexts**

    Mayadhar Sethy

*Discover Sustainability* ·
    [10.1007/s43621-026-04645-0](https://doi.org/10.1007/s43621-026-04645-0)

[VerusCite report](https://veruscite.com/share/4LrRd83BusEo3Hr3yOCUfm9rrDR6v43WtAugzO1JK2o)

### August 2026

**Large language models as judges: recent advances in LLM-based evaluation, critique, preference modeling, and feedback for text and code**

    Mihai Dan Nadăş

*Artificial Intelligence Review* ·
    [10.1007/s10462-026-11652-0](https://doi.org/10.1007/s10462-026-11652-0)

[VerusCite report](https://veruscite.com/share/OotE2MhGFZMtuMHbtZeDaOYAO5e0ew3F0PDwKQY2GkI)

**What students ask matters: LLM interaction depth, task quality, and immediate recall in higher education**

    Vangelis Tsiligkiris

*International Journal of Educational Technology in Higher Education* ·
    [10.1186/s41239-026-00617-6](https://doi.org/10.1186/s41239-026-00617-6)

[VerusCite report](https://veruscite.com/share/96c5S4IkcuhwBJxa8YjRDbkkXYtzfgS378JGigzWgzw)

**Generative AI bias against scientific novelty: a cautionary tale from a small-sample evaluation of research proposals**

    Diogo Machado

*Scientometrics* ·
    [10.1007/s11192-026-05792-x](https://doi.org/10.1007/s11192-026-05792-x)

**Spatiotemporal analysis of carjacking trends in Chicago before, during, and after COVID-19 lockdowns: implications for organized crime**

    Frank Gyebi Asiamah

*Trends in Organized Crime* ·
    [10.3390/ijgi10020005](https://doi.org/10.3390/ijgi10020005)

[VerusCite report](https://veruscite.com/share/J8aveLbZxxetxVOSSMiSRf8vSbiMAWStm8Q7gxBilA0)

**From crime to search engine: how crime predicts the search for security products?**

    Carlos Arango-Pastrana, Edwin Arango-Espinal & Carlos Osorio-Andrade

*Security Journal* ·
    [10.1057/s41284-026-00566-9](https://doi.org/10.1057/s41284-026-00566-9)

[VerusCite report](https://veruscite.com/share/IyRHJw60RkzlIfCqnKBL5NAWUB3SCqK1XnoA-zop4dA)

**Intersecting Generational, Gender, and Racial Disparities in Intimate Partner Physical and Sexual Violence: A Multivariate Analysis of Victimization and Perpetration**

    Sunday B. Fakunmoju

*Archives of Sexual Behavior* ·
    [10.1007/s10508-026-03494-2](https://doi.org/10.1007/s10508-026-03494-2)

[VerusCite report](https://veruscite.com/share/l42cFlsAt0gGysjNCqm9TcnPyIVe8iiLM42PfTzdAfI)

**Towards a Problem-Oriented Domain Adaptation Framework for Machine Learning**

*Information Systems Frontiers* ·
    [10.1007/s10796-026-10799-z](https://doi.org/10.1007/s10796-026-10799-z)

[VerusCite report](https://veruscite.com/share/gdU6HZuo8YgHCrlsDQZbxhO_K8ZbefQkwiIQtixM7bU)

    The
    [hall of hallucinations](https://veruscite.com/hallucinations)
    has recent articles across the major publishers — Taylor & Francis, Sage, Oxford,
    Elsevier, MDPI, PLOS One (I am currently missing Wiley!), as well as various preprint servers
    (arXiv, PsyArXiv, Frontiers, SSRN).

Having been involved with the process as an academic, I get it: publishers only do the minimal amount necessary and offload most integrity verification onto peer reviewers and editors. Given the margins these publishers command, though, academics should demand more from them in verifying the integrity of the work they publish.

My tool costs $2 to review a paper. If checking the accuracy of the references is not even worth that much to the journal, it does not seem to me there is much point in trusting anything peer review puts out at all.