Methodology: how Face Search actually works

The short version

When you run a paid search, your photo goes through a short pipeline: a face is detected in your image, converted into a numeric representation, and compared against a large index of publicly sourced photos for visually similar faces. You get back a ranked list of candidate sources to review yourself.

What powers the search — and why we tell you

Face Search does not build or maintain its own facial recognition index. Every paid search is executed by FaceCheck.id, a specialist provider whose core product is large-scale reverse face search. We integrate with their API to run the underlying match.

We think this is worth stating plainly rather than burying it. A lot of face-search products imply a proprietary technology moat that doesn’t exist. Ours doesn’t either, and pretending otherwise wouldn’t make the product better — it would just make our claims less trustworthy.

What Face Search adds on top of the raw API

If the matching itself comes from a partner, what do you actually get from us? Three things:

• A simpler, guided upload flow. Photo quality tips, a guest preview before you pay, and a results view built for someone who isn’t a professional investigator.
• Pay-once credits instead of a subscription. See pricing — Essential, Plus, and Ultra packs, credits that never expire, and a 30-day money-back guarantee.
• Honest guidance content. Guides on reading match scores, spotting catfishing patterns, and understanding legal and ethical boundaries — the parts a raw API doesn’t explain for you.

Step by step: what happens during a search

1. You upload a photo. We check basic quality signals (a single, front-facing, unobstructed face works best — see our photo-quality guidance).
2. A face is detected and encoded. The system locates the face in the frame and derives a compact representation used for comparison — not the raw image itself.
3. The encoding is compared against the index. FaceCheck.id searches its index of publicly sourced photos for visually similar faces and returns candidates with a relative confidence ranking.
4. You get a results page. Each candidate links back to its source (a page, profile, or article) so you can open it and judge context yourself — we don’t auto-resolve a match into a name or profile for you.
5. A credit is deducted. One credit is used per completed paid search, drawn from whichever pack you purchased.

What a match score means (and doesn’t)

Results come back with a relative confidence signal, not a certainty percentage about who someone is. Higher scores mean stronger visual similarity between faces; they do not mean the tool has confirmed a legal identity. Use the explainer below to see how we recommend treating different score ranges.

Known limitations — stated plainly

We would rather you know this before you spend a credit than after:

• Zero results is a valid outcome. If a face isn’t present in publicly crawled sources, there is nothing to return — this can happen even with a clear, recent photo.
• AI-generated and heavily edited faces behave unpredictably. Synthetic faces have no real photographic history to match against, so searches on them often return nothing, or return visually similar but unrelated people.
• Photo quality matters a lot. Side angles, heavy filters, sunglasses, low resolution, and group photos all reduce the chance of a useful result.
• Coverage is limited to what has been crawled. We don’t know, and won’t claim to know, an exact size or completeness figure for the underlying index — treat every result as a partial, not exhaustive, view of the public web.
• A match is a starting point, not a verdict. Always corroborate a result against other information before acting on it.

How your photo is handled

Your uploaded photo is used to run the search and is not used to train recognition models. For the full breakdown of data handling, retention, and your rights, read our ethics & privacy page and privacy policy.

Keeping this page accurate

We review this methodology whenever the underlying search process changes, and at minimum once a year. If you spot something here that seems inaccurate or out of date, email team@facesearch.id and we’ll correct it.