Referom — Stop disinformation at the citation.

Why Referom exists — plagiarism detection isn't enough anymore.

Citations and quotes in scholarly work are hard to verify quickly. Generative AI invents plausible-looking sources at scale, and the fixes show up months after the damage is done. Plagiarism tools cover text overlap — no one is really focused on referential integrity: checking that a citation points to a real source, that the source actually says what the paper claims, and that you can score and audit the whole thing.

Referential integrity, not just text overlap.

Referom pulls citations out of a paper, finds the original source for each one, and checks how well the citation holds up — does the source exist, does it actually support the claim, is the quote accurate — then produces a score you can audit. Every check is logged, so the result is reproducible.

Four academic audiences. One value: speed to quality.

1. Universities — research offices, deans of research, graduate schools, departmental chairs, writing-center librarians. Used on theses, dissertations, and grant deliverables.
2. High schools — school and district administrators, curriculum offices, learning-platform owners. Used on AI-assisted student writing where citation hygiene has degraded fastest.
3. Scholarly journals — editorial offices, managing editors, peer-review coordinators, manuscript-tracking-system owners. Used to pre-screen submissions before reviewer time is committed.
4. Conferences — program committee chairs, area chairs, conference organizers running tight peer-review cycles. Used to triage hallucinated bibliographies inside the review window.

API-first by design. Your workflow doesn't move — Referom plugs into it.

The web interface is one entry point. Everything also runs through an API, so it plugs into the tools you already have: LMS, research-information systems, manuscript-tracking systems, peer-review platforms. The integrity score, per-citation detail, and link back to the original source show up in the tool your team already uses.

Three guarantees on every analysis.

• Per-Citation Integrity Score — every citation gets its own 0–100 score with sub-scores for whether the source supports the claim, whether the paraphrase is faithful, and whether the right paper was attributed.
• Source Traceback — every claim links to the exact sentence in the actual source document, not just a DOI.
• Reproducible Audit — every check is stamped with the version of the system that produced it. Re-runs add to the record; old scores stay on the record, unaltered.

What Referom is not.

• Not a writing-style critique tool. Referom does not assess prose quality, scholarly tone, or venue fit.
• Not a plagiarism rating. Referom does not score text overlap against prior work. Plagiarism detection is a mature category with capable incumbents; Referom does not market or benchmark against them.

Three commitments before any line of code.

1. Integrity is measured, not asserted. Every score is reproducible and auditable.
2. The claim must reach the sentence. A citation is a claim about a specific passage in a specific source. The check isn't done until we've found that passage and compared the claim to what the source actually says.
3. Provenance is part of the score. Every check is stamped with the method that produced it. Nothing rewrites history.

Principal Investigator — Andrew Schwabe

Founder and author. Serial entrepreneur; previously founded ForMotiv, Saigon AI, and QCoda. Affiliated with the University of St Andrews. Building Referom because generative AI is flooding scholarship with content the world can't keep up with.

Join the Research Partner Program.

A small group of universities, schools, journals, and conferences is helping shape v1 before broader availability. Write to [email protected] for a direct introduction.