AI Ref-Assist

RefAssist: Editorial-Grade Reference Verification for IEEE-Style Publishing

Citation errors and fabricated references are a growing risk in academic publishing. LLM-assisted manuscript preparation has made invented citations and silently altered references far more common, while retracted work continues to slip through peer review undetected. RefAssist gives editorial teams an automated first line of defense.

What it does

RefAssist validates every reference in a manuscript against 14 authoritative bibliographic sources — including Crossref, IEEE Xplore, PubMed, arXiv, DataCite, and Semantic Scholar — before reformatting it to IEEE Editorial Style Manual rules.

• Identity verification, not just formatting. Matches each citation to the real underlying work using title, author, and identifier (DOI/ISBN) evidence, so look-alike titles and near-duplicate records are never silently merged.
• Fabrication detection. Flags citations with no matching record, journals that don't exist, dead DOIs, and DOI collisions — surfacing likely invented or AI-hallucinated references before they reach print.
• Retraction screening. Cross-checks every reference against Retraction Watch, Crossref, and OpenAlex retraction data, with a prominent notice and reasons on record.
• Field-level provenance. Every correction shows exactly which source confirmed it and a confidence rating — no black-box edits.
• Full transparency report. A per-reference audit trail: sources consulted, disagreements between databases, unconfirmed fields, and action items for the author or editor to resolve.
• Broad reference-type coverage. Journal articles, conference papers, books, book chapters, theses, technical reports, standards, preprints, datasets, and software citations.
• Batch processing. Paste a reference list or upload manuscript files (.docx, .pdf, .tex, .bbl, .txt) and process an entire bibliography at once.
• Flexible export. IEEE-formatted reference list, BibTeX, CSL-JSON, and a full audit report.

Built for editorial workflows

RefAssist reduces the manual burden on editorial and production staff by catching the errors that matter most before a manuscript goes to peer review or print — protecting your journal's accuracy and integrity at scale.