Verification Methodology
Defensible decisions. Minimum evidence.
Source-first · Separate confidence · Human gates
Bonafide is designed to answer what the evidence supports, not to produce a single “AI fake score.” Source confirmation, document consistency, regulatory status, identity consistency, channel trust and input quality remain separate dimensions.
Decision state machineNo universal fake score
Authority VerifiedSource evidence consistent
Highly ConsistentSupporting evidence strong
Evidence RequiredTargeted item can resolve uncertainty
Manual VerificationOfficial-channel confirmation needed
Potential DiscrepancyMandatory human review
Unable to DetermineNo fraud conclusion
Source-firstAuthoritative routes outrank appearanceMinimum evidenceEscalate only what can resolve uncertaintyHuman gateSerious adverse outcomes require reviewAuditableMaterial actions preserve lineage
WHAT BONAFIDE DOES
Purpose-built verification capabilities.
Each capability is a visible part of the verification workflow—not a hidden AI score.
◇Document authenticity confidence
Forensic/template/logical consistency.
✓Source verification confidence
Strength of authoritative source confirmation.
◎Regulatory confidence
Quality of recognition/status evidence.
↗Identity consistency confidence
Cross-document identity evidence.
▦Channel trust confidence
Trustworthiness of the verification response channel.
⌁Quality confidence
Whether scan/input quality supports the claimed analysis.
EVIDENCE MODELCustomer-facing states stay explicit.
The state machine can return AUTHORITY_VERIFIED, CRYPTOGRAPHICALLY_VERIFIED, HIGHLY_CONSISTENT, ADDITIONAL_EVIDENCE_REQUIRED, MANUAL_VERIFICATION_REQUIRED, POTENTIAL_MATERIAL_DISCREPANCY or UNABLE_TO_DETERMINE. MATERIAL_DISCREPANCY_CONFIRMED is gated by the required human-review policy.
- Minimum evidence first: ask only for the item needed to reach a defensible result
- Immutable originals: derivatives never replace raw evidence
- Temporal truth: evaluate university/company/regulatory status for the relevant historical period
- Provenance everywhere: contacts, templates, regulatory records and manual confirmations carry source class/timestamp/confidence
- Do not overclaim: the product supplies evidence and verification states, not the customer’s admission/hiring/visa decision
Decision state machineNo universal fake score
Authority VerifiedSource evidence consistent
Highly ConsistentSupporting evidence strong
Evidence RequiredTargeted item can resolve uncertainty
Manual VerificationOfficial-channel confirmation needed
Potential DiscrepancyMandatory human review
Unable to DetermineNo fraud conclusion
WORKFLOW
From input to defensible evidence.
Bonafide follows the strongest available verification route and resumes only the affected lane when new evidence arrives.
01Receive evidence→
02Run independent dimensions→
03Apply policy→
04Escalate only if necessary→
05Human-confirmed final state
DESIGNED FOR
One evidence model. Role-specific operations.
The same underlying case can be viewed differently by institutions, employers, reviewers, operations and candidates.
01Reviewers
Explain why a result exists
02Enterprise customers
Configure policy without bypassing safety rules
03Candidates
See evidence states rather than accusations
DECISION DISCIPLINEUnknown ≠ fake.
Unmapped issuer, new template, legacy document or unavailable source must route to uncertainty/manual verification rather than an automatic accusation.
Authority VerifiedAuthoritative source evidence is consistent.Highly ConsistentStrong supporting evidence; independent source not complete.Evidence RequiredA targeted item can likely resolve the case.Unable to DetermineEvidence remains insufficient; no fraud conclusion.
DCS BONAFIDE · CREDENTIAL INTELLIGENCE PLATFORM
Establish what is bona fide.
Preserve the evidence.