What a selfie ID match tells a marketplace, what it cannot tell you, how to read the result and how to design around its limits.
Ask a seller to photograph their ID and take a selfie, and most people understand why. The face on the document should match the person signing up. It is one of the most familiar steps in online onboarding, and one of the most misunderstood. Teams often assume a selfie check proves more than it does, and design their risk rules around that assumption.
This post explains what selfie-to-document matching actually confirms, what it does not, how to read TrustVerify's result, and how to build an onboarding flow that is honest about the limits.
The idea is simple. The seller provides two images: one of their identity document, which includes a photo, and a selfie. The system compares the face in the selfie with the face on the document and assesses whether they are likely to be the same person.
TrustVerify's selfie checks guide puts it plainly: we compare the selfie with the photo on the ID document, so you know the face on the ID matches the person who took the selfie. In the API response, the face match is reported under signals.biometrics.faceMatch, and it is assessed on every identity check that includes a selfie. The value is true, false or null. Behind it, the two faces are compared and given a similarity score from 0 to 100; by default a score of 50 or more counts as a match. A score below that sends the check to review rather than rejecting the seller, and a very low score is flagged as a face mismatch, which is still a review, never an automatic rejection.
Combined with the document check, that gives you two things:
That is valuable. It stops the simplest form of identity misuse, where someone uploads another person's real ID with their own face, or a random photo, in the selfie.
Being precise about the limits matters more than the strengths, because the limits are where risk rules go wrong.
It does not confirm how the selfie was captured. A face match compares two images. It does not tell you whether the selfie was a fresh photo, a photo of a photo, a screen, or a manipulated image. Treat it as a comparison, and cover capture risk with your other checks.
It does not confirm the person is honest. A genuine seller with a genuine document can still intend to defraud buyers. Face matching says nothing about intent.
It does not confirm the document belongs to the person in law. It confirms a visual resemblance between two faces, not a legal link established some other way.
It does not confirm the rest of the application. Address, email, phone and business details all need their own checks.
None of this makes face matching weak. It makes it one part of a larger set of checks.
TrustVerify's guidance is explicit: a null signal means no result was returned for that field, so never read it as a pass.
This is a common bug. Code that checks if (faceMatch !== false) will treat null as success. Code that checks if (faceMatch === true) will not. Write your rules so that only a positive result counts as a pass, and treat null as "not assessed".
The guide also gives the practical answer for teams whose risk policy needs more than a face match: hold those sellers in review or add a step of your own.
Face comparison produces an assessment of how likely two faces are to be the same person. Like any comparison, it can be wrong in both directions:
Because errors happen in both directions, a non-match should not mean an automatic permanent rejection. Give the seller another attempt with clear guidance, and send repeated non-matches to a person.
Many non-matches are photo problems. The selfie guide recommends:
For the document image, the capture guide recommends JPEG or PNG, around 1,500 to 2,000 pixels on the long edge, the whole document flat in frame and no glare over the photo.
Small interface touches help. Show an oval guide for the face. Tell people to remove glasses if the first attempt fails. Show the captured selfie before sending it. These cut retakes and reduce how often genuine sellers end up in review.
If face matching confirms resemblance but not how the selfie was captured or what the seller intends, the rest of your flow should cover the gaps in proportion to the risk.
Combine signals. Sanctions and PEP screening, email and phone checks, and device fingerprinting with bot detection each cover something face matching does not. TrustVerify's fraud score combines device, IP and bot signals into an approve, review or block outcome, explained in the fraud score guide.
Use review for uncertain cases. Non-matches, nulls where your policy needs a result, and odd combinations of signals should go to a person. The review queues page shows how.
Limit before you trust. A newly verified seller can start with lower payout or listing limits that rise with a clean track record. That reduces the damage if something slipped through.
Watch for change. A face match at onboarding says nothing about who is using the account six months later. Changes of device, email or payout details are worth watching. Our planned post on marketplace account takeover covers this. [Planned URL, live on publication.]
Two kinds of honesty matter here.
With sellers, explain what the selfie is for, how it is used and how long it is kept. Facial images processed to confirm identity can be special category biometric data under UK GDPR, so your privacy information and lawful basis need care. The ICO's biometric recognition guidance is the place to start. This is general information, not legal advice.
In your own materials, describe the check accurately. "We match your selfie to your ID" is accurate. "We confirm who you are beyond doubt" is not. Overstating checks to partners or users creates a gap between what people think you do and what you do.
faceMatch as true, false or null on identity checks with a selfie, and null is never a pass.Selfie matching is one part of our KYC software for UK businesses; to compare providers, read our guide to choosing KYC software in the UK.
See how selfie matching fits with document checks and screening in seller onboarding. Read the selfie checks guide.
KYC, KYB, AML screening and fraud prevention in one place, with published pricing.
KYC
Passport MRZ and Selfie Checks Explained for Marketplace Teams
What a passport MRZ is, how its check digits work, what a selfie-to-document match confirms, and what neither check proves on its own.
KYC
Documents a Marketplace Can Accept for KYC, and How to Handle the Rest
Which ID documents marketplaces commonly accept for seller KYC, what to do with documents outside that list, and how to get clear images first time.
KYC
Proof of Address Checks: What They Prove and What They Don't
What a proof of address document actually shows, its limits for marketplace seller checks, and how to use it alongside stronger signals.