Pilot program

Research Integrity, While You Write

Citation and reference checks inside SciFlow

Retracted sources. Invalid DOIs. References that do not exist. Uncited literature. Conspicuous citation patterns.

SciFlow checks all of this while the manuscript is being written – before it reaches an editorial office, a reviewer, or a publisher.

The problem

Problems are often found only when it is too late

A reference can look plausible while you write and still be a problem:

  • The cited work has been retracted.
  • The DOI leads nowhere.
  • The bibliographic details do not match.
  • A reference was produced by an AI tool and does not exist.
  • An entry in the bibliography is never cited in the manuscript.

Today such problems usually surface during editorial screening. SciFlow moves that check to where the references are created: into the writing process.

What is a quick correction while writing can later mean a query, a revision, or a delay.

The checks

Direct analysis of the references

01

Verify sources

Can the cited source be identified unambiguously? Do the title, authors, DOI and other bibliographic details match?

02

Detect retractions

Has a cited work been retracted, or carries a corresponding editorial notice?

03

Check reference relevance

Does the source fit the manuscript and the passage that cites it?

04

Reconcile the bibliography

Which entries in the bibliography are actually cited in the manuscript?

05

Analyze citation patterns

How are self-citations and citations of particular authors distributed across the manuscript?

Why in SciFlow

The check happens where the reference is created

The manuscript and its bibliography are already in SciFlow. The check runs without a change of tools – no export, no upload – and every finding leads back to the place in the document where it is corrected.

For the analysis we work with Grounded AI. Their Veracity is specialized in checking scholarly citations and, according to the vendor, already in use at academic publishers.

Data protection

Your manuscripts stay in SciFlow

For most checks the bibliography is enough. The manuscript text does not have to leave SciFlow.

When full-text analysis is enabled, only the document version needed for that check is transmitted. The analysis is carried out by Grounded AI under a data processing agreement with SciFlow. The transmitted data is processed solely to produce the citation report and is deleted afterwards.

Bibliography

Source resolution, DOI validation, retractions and bibliographic metadata need nothing but the reference list. The manuscript text stays in SciFlow.

Full text

Relevance analysis and the reconciliation between citations and bibliography additionally require the manuscript text.

You decide the scope of the analysis before every run.

From the check to the corrected manuscript

  1. 01Choose the scopeBibliography or full text.
  2. 02Start the checkSciFlow analyzes a specific version of your document.
  3. 03Review the findingsConspicuous references and citation patterns are shown in the Research Integrity Center.
  4. 04Correct the manuscriptEvery finding leads back to the corresponding place in the document.

The report tells you where to look. The correction happens in the manuscript itself.

Pilot program

Become a pilot partner

The Research Integrity Center is currently in pilot operation.

We are looking for institutions that want to bring citation and reference checks into their academic writing workflow early, and develop the feature further together with us. As a pilot partner you get:

  • access to the current version of the checks
  • an introduction for your team
  • the option to test the feature for an entire institution or for selected user groups
  • direct influence on how it develops
Request pilot access

Request pilot access

Tell us who you are and what you want to check. We will explain what the pilot can and cannot currently do for your institution.

What the documentation covers

The support article documents the current pilot feature set: what each check does, how the two analysis scopes differ, and how the data is processed.

Read the documentation