Integrity
Signals, not silent verdicts.
Automated checks never change a score. They attach evidence to the record so the committee can judge it, cite it and defend it later.
Text similarity
Essays are compared against prior submissions and public sources. The reviewer sees the matched passage, not just a percentage.
Fraud & anomaly
Duplicate identities, recycled references, implausible timelines and inconsistent financial claims raise structured signals.
Document authenticity
Transcripts and certificates pass type, tamper and third-party verification checks where connectors exist.
Evidence gaps
Claims made in the essay without supporting evidence in the file are flagged for follow-up, not penalised automatically.
Score drift
Reviewer severity and leniency are measured across the cohort so chairs can calibrate instead of guessing.
Immutable audit
Access, edits, overrides and decisions append to a log that cannot be rewritten.
Human judgement is final. AI output is a suggestion with citations; the accept, edit or override action is recorded against the person who made it.