How NoveltyAxis checks its work

AI tools can sound certain about things that are not true. NoveltyAxis is built so that a report can only say what it can point to. This page explains how.

Create a free account

Quotes are checked word for word

Whenever a report quotes a paper, the quote is matched against the text of the paper that was actually retrieved. A quote that cannot be found is marked as not found, and claims that rest on it are not presented as evidence.

In paper reviews, every point on the fix list quotes your own draft, so you can find the passage and judge the point yourself.

Every statement says what kind it is

  • Evidence: a verified quote from a retrieved paper.
  • Inference: a conclusion drawn from cited evidence.
  • Hypothesis: proposed, not yet tested.
  • Assumption: taken as given, without evidence.
  • Experimental result: observed in an approved code run.

References are checked against public records

The reference checker looks every entry up in Crossref, OpenAlex and arXiv and compares the title, year and authors. It flags entries that do not exist, point to a different paper, or have been retracted. This check uses no language model.

Reports are built from recorded data

The final report is assembled from what the run recorded, not written from memory, so it cannot cite a paper that was never retrieved. Journal quartiles come only from the SCImago file loaded on the server and are never guessed.

What we do not claim

  • That an idea is novel. We show the closest published work we found and how your idea differs.
  • That a gap is certain. We show the evidence behind it and how recent it is.
  • That a review guarantees acceptance.

Questions

Which sources does NoveltyAxis search?

arXiv, OpenAlex, Semantic Scholar and Crossref: open scholarly indexes that cover every field.

Is my work used to train AI?

No. Your idea or draft is used only to produce your report.