KERNIT Documentation

Review Workflow

The safe KERNIT workflow is scan first, resolve DOI evidence only where it helps, review risky matches, then apply links with the same file and settings used during scan.

Why Review Comes Before Apply#

Cross-reference automation can create wrong links if a bibliography is duplicated, a citation is ambiguous, or a figure/table label was edited manually. KERNIT separates detection from final writing so reviewers can correct the target before the DOCX is changed.

Decision Types#

DecisionMeaningWhen to use it
KeepUse the suggested target.The citation or cross-reference points to the correct entry.
SkipDo not write a hyperlink for this match.The match is ambiguous, decorative, or intentionally unlinked.
RetargetChoose a different target.The text is real but KERNIT selected the wrong bibliography entry, caption, or section.
Link onlyWrite URL/DOI evidence without changing citation identity.Resolver evidence is useful but identity-changing text edits should be held.
Important: apply must use the same DOCX and settings used during scan. The API enforces this with scan tokens and fingerprint checks.
  1. Scan the source DOCX

    Collect review rows and target choices without modifying the file.

  2. Resolve references

    Use Crossref evidence for DOI and title matches when the bibliography needs external metadata.

  3. Review uncertain rows

    Prioritize unmatched, low-confidence, duplicate-author, same-year, and identity-changing matches.

  4. Apply the approved decisions

    Submit the same file, same settings, scan token, and decisions to produce the linked DOCX.

Was this page useful?Send a lightweight docs feedback event.
Last updated: May 12, 2026Canonical: https://kernit.org/docs/getting-started/workflow/