What is a reverse face search?

What does reverse mean here?

In an ordinary search you type words and get pages. In a reverse search you start from the thing you would normally be looking for, a picture, and ask where it exists. A reverse image search starts from a whole picture and finds copies of it. A reverse face search starts from the face inside a picture and finds other pictures that contain the same face. The word reverse only says that the input is an image rather than text.

That small difference in input changes what comes back. Copies of one picture are useful when you want to know where a specific photo was reposted. Other pictures of the same person are useful when you want to know where that person appears at all: a conference gallery from years ago, a profile under another name, a news article, a forum avatar. The two kinds of search answer different questions, which is why face search vs reverse image search is worth reading next.

How does a reverse face search work?

Every engine in this category runs roughly the same pipeline, with its own models and its own index. The steps below are the generic version; the FaceSearch version is on how it works.

1. 01Detection. The engine finds the face in your photo and crops to it. If it cannot find a face, or finds several, it asks you to pick or to upload a clearer picture.
2. 02Alignment and normalisation. The crop is rotated and scaled so the eyes, nose and mouth sit where the model expects them. This is why a tilted or turned face still works, up to a point.
3. 03Embedding. A neural network turns the aligned face into a list of a few hundred numbers. Two photos of the same person, taken years apart in different light, produce lists that sit close together. Photos of different people sit further apart. The list is often called a face embedding, a face template or a feature vector; they all mean the same thing.
4. 04Comparison. The engine has already computed embeddings for faces it found while crawling public web pages. It measures the distance between your embedding and each stored one.
5. 05Ranking. Pages whose faces are closest come back first. Most engines show the matched image, the page it came from and some indication of how close the match is.

Nothing in that pipeline knows who anyone is. The engine compares shapes, not identities. A result is a public page where a similar face appears, which is a lead a human has to confirm. The doppelgänger effect explains why close is not the same as identical.

What does a reverse face search find?

It finds faces on publicly accessible web pages that the engine has crawled. In practice that covers a large and useful slice of the web: news sites, blogs, company pages, event and photography galleries, forums, public profiles that search engines can also see, product listings, and the many sites that repost pictures from elsewhere. For most people a self-search surfaces photos they had forgotten about, photos they never knew were public, and sometimes photos used without their consent.

• Different photos, same person. The defining capability. One current photo can find pictures taken a decade ago.
• The same photo, edited. Crops, flips, filters, added text and lower resolution rarely stop a match as long as the face is still clear.
• The source page. Each match comes with the URL it was found on. That is what turns a result into something you can act on: a name to check, a site to contact, a report to file.
• Context. The page itself tells you the when and where: a dated article, a gallery title, a username. The face match gets you to the page; the page tells you what it means.

What can it not do?

Who uses reverse face search, and for what?

The most common search is a self-search: someone wants to know where their own face appears. From there the uses branch. A person on a dating app wants to know whether the photos they are looking at belong to the profile that posted them; see the catfish check. A creator wants to know whether their content is being reposted; see face search for creators. Someone who has found a fake account using their photos wants every copy before they report it; see fake profile impersonation. Journalists and researchers use it to check whether a photo attached to a claim has appeared before under another story; see journalists and researchers.

The same tool can be misused, which is why every reputable engine publishes acceptable-use rules and a removal route. Stalking, harassment, outing someone, and screening people for jobs, housing or credit are the standard prohibitions. Ours are on the acceptable use page.

Is reverse face search the same as facial recognition?

They share the underlying technology and differ in purpose. Facial recognition usually means identifying a person: matching a face against a database of known people and returning a name, as a phone does when it unlocks or as a border system does against passport records. Reverse face search matches a face against faces found on public web pages and returns those pages. There is no enrolled database of named people and no identity decision. The distinction matters legally as well as technically, which is why the facial recognition search page and the legal guide treat it carefully.

Is reverse face search legal?

Searching for your own face is lawful in every market we serve. Searching for someone else's face is lawful for ordinary purposes in most places, but biometric and data-protection laws vary a great deal: the EU treats face data as a special category under the GDPR, Illinois has a dedicated biometric privacy statute, and India's Digital Personal Data Protection Act is still being implemented. Some uses, such as employment or tenant screening, are restricted or prohibited regardless of jurisdiction. The country-by-country detail is in is face search legal?.

How this applies to FaceSearch

FaceSearch is a reverse face search engine in exactly the sense above: one photo in, matching public pages out, with no names and no database of people. Your first search is free with no account. The photo's metadata is stripped before anything else happens, the upload is used only to run your search, and it is never sold or used to build a face database. A deep search widens the pass over the public web when a normal search comes back thin. If you find a photo of yourself that should not be public, the removal guide covers what to do, and the remove my photos page explains how to take your face out of our results.