Report

Enforcing the requirement to cite sources in AI research through technical means

For literature and evidence review in regulatory studies, we have developed a pipeline that does not allow any source to pass through without a full text, nor any finding without a page anchor.

Let’s discuss your project

Contact us directly and let's talk about a possible collaboration.

Contact

Background

AI-powered research quickly returns literature, including sources that do not exist and page references that do not support the cited statement. In regulatory proceedings and scientific publications, even a single such error can prove costly. The only reliable option is a collection that is stored locally and has been verified.

Our Approach

  • 1

    Retrieval and local storage of each source prior to analysis, with search procedures logged and any failures documented.

  • 2

    Mapping of each key finding, including page references, quality assessment and project relevance, in an ongoing source map.

  • 3

    Automated systems that reject a source lacking full text and a finding without a verifiable anchor before citing from it.

Result

The end result is a citable local catalogue: no entry in the deliverable is complete without the page on which it appears. The first automated inventory of an established catalogue revealed serious errors that nobody had spotted before; following the changeover, the same catalogue passes through the system without any issues. Conflicts between sources and missing references are listed in the source map, rather than disappearing into the text.

contact