Published, auditable methodology

Every confidence score must explain itself

Confidence is a deterministic sum of signed weights, clamped from 0 to 100. It prioritizes review and never auto-publishes verification.

Signed confidence weights

Positive signals

Official site links account+35

Confirmed official cross-link+25

Platform verification indicator+20

Institutional page links account+20

Representative or management page+15

Consistent website domain+10

Consistent bio, profession, and country+8

Reputable secondary source+5

Historical continuity+5

Negative signals

Official source links another account-40

Fan, parody, or unofficial label-35

Public impersonation warning-35

Conflicting website ownership-30

Identity or profession conflict-25

Handle spoofing-20

Conflicting cross-links-20

Strong duplicate conflict-15

Stale or inaccessible primary evidence-10

Name-only similarity = 0 points

Text similarity never raises confidence and cannot connect an account to a person. An independent, reviewable source is required.

Tiers, risk, and priority

Confidence tiers

High confidence90–100Likely75–89Unconfirmed50–74Weak25–49Excluded0–24

Risk weights

Low+10Medium+25High+45Critical+70

Review priority

priority = clamp(risk×0.45 + uncertainty×0.6 + reports×4 + changed×18 + watchlist×10 + conflicts×6, 0, 100)

Follower count is absent from the formula and never determines order.

Current registry and sources

180,314Profiles

378,620Accounts

640,973Evidence

62Countries

WikidataWIKIDATA565,625

Official websiteOFFICIAL WEBSITE19,189

Cross-platform handle consistencyPUBLIC PAGE CHECK17,590

Official website scanOFFICIAL WEBSITE15,287

TikTok oEmbedPLATFORM API

Privacy boundary

IKNUVA separates the public registry from account and claim data. Sign-in email is used to manage and verify the account; registering or submitting a claim does not publish that email on a public profile.

Claims require account links and a private photo accessible to the claimant and authorized reviewers. Contact numbers are optional and private by default; publication requires separate, explicit claimant consent and claim approval. Submitting a photo or link does not publish it automatically or prove ownership on its own.

In the picture-verification tool, the downscaled picture is sent only with explicit consent and only for that check: its fingerprint is computed in memory, compared with registry pictures and discarded, never stored. The face descriptor is computed in the browser and compared with public registry profiles only; we do not search faces on the open web. A separate face-matching service receives a numeric descriptor and compares it with an image-descriptor index; its matching endpoint does not receive the visitor image or store the lookup descriptor. A face match does not prove identity or ownership.

When submitting a report or evidence, use public links and do not attach private messages, government identification, or financial or medical data. These boundaries do not replace reviewing the information you choose to submit.

Public names and handles with no search results may become discovery requests after filtering sensitive-looking text. The name may be sent to Wikidata and then YouTube to find candidates, without the visitor's IP address or account information. Name similarity does not prove identity.

When the YouTube source is enabled, we use YouTube API Services to discover public channels. We retain channel names, links and public-link suggestions, refreshing or removing temporary copies before 30 days. Request removal through the profile's report form; removing our copy does not delete data from YouTube. IKNUVA scores and trust decisions belong to our product, not YouTube, and a channel's existence does not establish a person's ownership.

YouTube Terms of Service Google Privacy Policy Google permissions

Open evidence from each profile and submit new information for human review.

Explore the registry