Skip to main content
Verification · step 5

Three machine passes. A person still signs Verified.

Licence cards hit the public registry for that country. Sterilization files trigger a search of the dentist and the practice on official cert lists. Case photos are reverse-searched on the open web and indexed social. None of this publishes a dentist.

Where it sits

On every upload — not only photos

Licence → public registry

OCR the card, then query the issuing body for that country (SEP cédula, TDB, ReTHUS, GDC, DHA…). Active and unrestricted, or pillar 1 stays closed. A clinic permit is not the dentist.

Sterilization → name + practice

A yes on the application is not a pass. We search the dentist and the practice on JCI, ISO 17665, GHA, and national CSSD / sanitary lists. They still owe the walkthrough video.

Cases → reverse image

Perceptual hash against stock, the open web, indexed Instagram / Facebook / TikTok, and every pair already on RatedDocs. Private accounts and Stories are not searchable — we do not pretend they are.

The seven checks

What the machine looks at

  1. 01

    Ingest

    Normalize crop, color, and metadata. Missing camera EXIF is a signal, not a verdict — phones strip it all the time.

  2. 02

    Perceptual hash

    A fingerprint that survives resize and mild compression. Matches our stock and prior-upload index.

  3. 03

    Web reverse lookup

    The same pixels on clinic sites, directories, and image CDNs — including other dentists’ pages.

  4. 04

    Social graph

    Instagram, Facebook, and TikTok posts where the photo already ran as an ad or someone else’s case.

  5. 05

    RatedDocs corpus

    Every case file already submitted on RatedDocs. A hit here means another dentist already claimed the work.

  6. 06

    Synthesis signals

    Texture, gum geometry, and generator fingerprints. High score holds the file for a reviewer — it is not an automatic reject.

  7. 07

    Pair identity

    Before and after should be the same person, same camera distance, plausible time gap. A swapped face fails this check.

Licence lab

Query the board, not the brochure

On upload we OCR the card and hit the public registry for that country. No public lookup means a reviewer writes the board — it is not an automatic pass.

  • Mexico. SEP cédula profesional + state colegio
  • Turkey. Türk Dişhekimleri Birliği
  • Colombia. ReTHUS (Talento Humano en Salud)
  • Peru. Colegio Odontológico del Perú
  • Costa Rica. Colegio de Cirujanos Dentistas
  • India. Dental Council of India + state dental council
  • United Arab Emirates. DHA / DOH / MOHAP (emirate of practice)
  • Hungary. Hungarian Medical Chamber (MOK) dental roll

Sterilization lab

Search the dentist and the practice

Ticking “yes, we have a protocol” never passes pillar 2. We search official cert lists under the dentist’s legal name and the practice name, then still require the walkthrough.

Teaching lab

Run four recorded cases

These are staged files so a team can see each check fire. Production Auto Scan hits live web, social, and RatedDocs indexes — this lab replays the outcome.

Before teaching case
Before
After teaching case
After

Dr. Morales · teaching case

Submitted as her own veneer case.

  1. Ingest

    Queued

    EXIF + normalize

  2. Perceptual hash

    Queued

    Stock + prior index

  3. Web reverse lookup

    Queued

    Open web + CDNs

  4. Social graph

    Queued

    IG / FB / TikTok

  5. RatedDocs corpus

    Queued

    Other dentist files

  6. Synthesis signals

    Queued

    AI fingerprints

  7. Pair identity

    Queued

    Same person?

Clear

No collision, low synthesis. Pillar 4 can proceed. A reviewer still looks at the pair.

Hold

Missing EXIF or a missing before frame. Not a reject — the dentist is asked for the raw file or the other half of the pair.

Flag

Stock, another dentist, or a generated smile. The listing can stay up; pillar 4 stays closed until the photo is replaced.

See it in the verification path

Auto Scan is step 5 of 6. Paying or uploading is not the same as being verified.