← Back to VOUCH

VOUCH trust, evidence and score rules

Updated 2 October 2026. Contact verification, claims, challenges and paid searches are not yet available on the website preview.

Verification and claiming your profile

The approved Verified Account standard is control of both a phone number and an email address. It does not certify legal identity, a person's character, or the truth of a relationship claim. Profile ownership and relationship verification are separate.

A notification link only routes a claim: it must expire after 24 hours, be single-use, and never grant ownership by itself. Phone verification and recycled/shared-number conflict checks are required. Matching names alone do not establish ownership. Competing claims must be reviewed; no automatic reassignment.

A wrong-number challenge does not automatically erase, detach or reassign a record. Its association and relationship remain Unverified and have no direct score effect. The closed case reopens if the original reporter corrects the number or the verified phone holder claims the profile and contests the relationship. Corrections preserve an audit trail.

Production contact verification and the complete claim journey remain launch blockers.

Evidence guidance

Use a message screenshot with a visible phone number, or an account conversation with its visible account identifier. A saved contact name alone is insufficient. An identifier helps establish which account was involved; it does not by itself prove who controlled that account or that a claim is true.

Provide relevant dates and enough surrounding conversation to understand the context. Keep the original available for private review. Disclose crops or annotations. Do not fabricate messages, alter their meaning, impersonate another person, or submit another person's experience as your own.

Duplicate submissions must not create extra score weight. Shared or recycled numbers and conflicting account identifiers require review rather than an automatic identity merge. Suspected falsification or impersonation requires review, appropriate exclusion/removal and a recorded decision.

Private materials and publication

Raw uploads, evidence metadata, full phone numbers, email addresses and legal identity details stay private. Purchased searches do not grant access to the evidence vault. Access is limited to the submitter and specifically authorized reviewers under the applicable permissions.

Only permitted profile information, an attributed experience or factual statement, its status, and an allowed response may be displayed through the appropriate access level after policy review. Evidence files are not automatically published.

Redact unnecessary identifiers, addresses, payment details, third-party names, children and unrelated messages from any approved display. Keep identifying context needed for review in the protected original, not a public screenshot. Intimate material, credentials and unrelated private information must not be published.

Understand the labels

- Opinion: the contributor's subjective experience or assessment.

- Reported fact: an attributed factual assertion that has not been established by the applicable review process.

- Verified fact: a specific fact confirmed through the applicable verification process; show what was verified and when. A Verified Account badge is not verification of every statement.

- Unconfirmed / Unverified: insufficient confirmation, including a one-sided relationship record.

- Disputed: a challenge is pending or a specific association or statement is contested.

Reported relationship status and dates must be shown with their verification state and relevant dates, rather than presented as guaranteed current facts.

How the Relationship Score is calculated

A score is an aggregate of eligible structured relationship reviews, not a truth finding or a guarantee. Individual 1–5 star assessments are separate from the 300–850 Relationship Score.

Only complete reviews accepted as score eligible enter the calculation. One-sided Unverified relationship records do not directly change the score. Excluded reviews and reviews under an approved technical calculation hold do not contribute.

Scored questionnaire answers use 0, 25, 50, 75 or 100; reverse-scored answers are inverted. Not-applicable answers are omitted. Category averages are combined using the active questionnaire's category weights.

Eligible reviews receive recency weights of 1, 0.5, 0.25 and so on, normalized across the included reviews. Ordering uses a verified relationship end date where available, otherwise the review submission/creation date.

Only verified relationship dates affect the duration multiplier: under 92 days = 0.75; 92–182 days = 0.90; 183–729 days = 1.00; 730+ days = 1.10. Missing or unverified dates use 1.00.

The weighted internal rating is converted once: rounded 300 + (rating × 5.5), bounded to 300–850. With no eligible reviews, display “Insufficient data” and no score, not a default score of 300.

Bands: 800–850 Exceptional; 740–799 Excellent; 680–739 Good; 620–679 Fair; 540–619 Needs Improvement; 300–539 Significant Concerns.

Review-count confidence is Developing for one review, Moderate for two or three, and High for four or more. These labels describe data quantity, not factual certainty. The active category weights and algorithm version must be accessible alongside the full methodology before launch.

Challenges, corrections and scores

The person discussed can respond and challenge identity, relationship, evidence, factual content, duplication/impersonation or score calculation. A factual challenge marks the specific claim Disputed; it does not automatically freeze the whole questionnaire score. An approved preliminary technical calculation hold excludes the affected review while reviewed.

Moderation outcomes must update the relevant visible label/content and, when eligibility, answers or verified dates change, recalculate the aggregate. Removed content remains unavailable to purchasers; purchased access does not promise access to removed material. Preserve the audit trail and notify affected people through the approved process.

The complete challenge, correction and upload journey still requires end-to-end testing before launch.

Access and misuse

Search access is for personal dating decisions. Automated searches, bulk harvesting, scraping, resale, shared access, impersonation, harassment and stalking are prohibited. Unlimited subscriptions do not authorize these uses.

Use authenticated, authorized endpoints and minimal pre-purchase information. Raw evidence and private identifiers must never be exportable through a purchased search. Temporary abuse restrictions require review and must not become an undisclosed ordinary-use daily search cap.

Server-enforced throttling, enumeration protections, audit monitoring and operational abuse response must be verified before search launch. Publishing this rule is not proof those controls are active.

Demonstrations and tests

Use visibly fictional profiles and dramatized stories. Do not seed claims about real people. Test fixtures must be clearly fictional and isolated or rolled back; screenshots must identify illustrative data.