Face search for investigators and fraud teams

What you are up against

Three recurring questions in fraud, due-diligence and open-source work are really one question: has this face been seen before, and where? Image matching answers it only when the file is unchanged. Scammers crop, mirror and filter precisely so that it is not.

• Recycled identitiesThe same headshot behind a new name on a marketplace, a loan application or a dating profile. The photo is real; the person presenting it is not the person in it. Finding the earlier appearances is what breaks the story.
• Synthetic and stolen profile photosFake accounts built from a real person's public photos, or from AI-generated faces that have no earlier appearance at all. A face search finds the former and comes back empty on the latter, and knowing which is which is itself a finding.
• Corroboration under time pressureA tip arrives with a screenshot and no provenance. The question is not who this is but whether the image has a history: a press photo, a stock portrait, an old profile. The search takes seconds; the corroboration takes the rest of the afternoon.

FaceSearch returns publicly accessible pages where a similar face appears, ranked by similarity, each with its source URL. It does not return names, identity records or anything from behind a login. The results guide explains why a close match is a strong lead and not a confirmation, and why look-alikes are a routine false positive in a large index.

Three workflows

1. Provenance check on a single photo

Crop to the face, search, and open every source page before drawing a conclusion. Note the earliest dated appearance and the context it sits in: a professional portfolio suggests a stolen stock-style image; an old personal profile under another name suggests identity theft; nothing at all suggests a new, private or AI-generated photo. Record the URLs and dates in the case file as you go, because a result list is evidence only while the pages still exist. The screenshot and cropping steps are on photo tips.

2. Reuse sweep across a set of profiles

When several accounts are suspected of being one operation, search the profile photo of each and look for overlap: the same face across accounts, or faces that trace back to the same portfolio. Keep each search to one face. Group photos match poorly and multiply the privacy exposure of bystanders. For sets of more than a few dozen, the Business plan's credit volume and deep search credits are what you need; there is no batch or API route, and every search runs through the interface.

3. Corroborate, then decide

A face match is one data point. Before it affects anyone, corroborate it with non-face evidence from the source pages: names, dates, locations, the account's own history, documents you already hold. The verification guide walks through the non-photo checks, and the fake profile playbook covers the reporting routes if the finding is impersonation. Decisions that affect someone's job, home, credit or insurance may not be based on a FaceSearch result at all, as the rules below explain.

Which plan fits

Lawful basis and client authorisation

Searching a face that is not your own is personal data processing. In the EU and UK you need a lawful basis under Article 6 of the GDPR (for most investigative work, legitimate interests weighed against the person's rights and documented before the search), and the biometric rules in Article 9 can apply to face data. In the United States, Illinois's Biometric Information Privacy Act and similar state laws restrict biometric processing, and the Fair Credit Reporting Act governs anything used for employment, housing, credit or insurance decisions. Get the client's written authorisation, record the purpose, and keep searches proportionate to it. The legal guide covers the US, EU, UK and India in plain language. It is not legal advice, and a licensed investigator should know the rules of their licence better than we do.

Rules that apply to you

• Never use a result for employment, tenant, credit or insurance decisions. FaceSearch is not a consumer reporting agency, and results do not meet the accuracy, notice and dispute requirements such decisions carry. Our acceptable use rules make this a condition of the account.
• Never search anyone under 18. The rule is a policy you agree to, not a technical block, and it applies to investigators as it does to everyone.
• Do not re-identify people who have chosen to be anonymous, track or surveil protesters, journalists or activists, or build dossiers on private individuals from match results.
• Keep searches to the matter you are authorised to work on, one face per search, and delete scans when the matter closes.
• Report what you find through proper channels: platform reports, the client's counsel, law enforcement. Do not contact the person in the photo.