Face search glossary

Face search borrows words from computer vision, from copyright law, from data protection law and from fraud prevention, and most pages use them as if everyone already knew them. This page defines each one in a sentence or two, in the sense FaceSearch uses it, and links to the page that goes further. Terms are grouped by topic rather than listed alphabetically, so that related words sit together.

What are the search terms?

Reverse face search

A search where the query is a face instead of words. You upload one photo, the engine finds the face in it and returns public web pages where a similar face appears. It matches the person, not the picture, so it can find different photos taken years apart. The full explanation is on what is a reverse face search.

Reverse image search

A search where the query is a whole picture. Google Lens, TinEye, Yandex and Bing compare the image itself, so they are good at finding copies and edited versions of one photo and poor at finding other photos of the same person. The two kinds of search complement each other; see face search vs reverse image search.

Normal search and normal credit

A normal search is FaceSearch's standard pass over the public web, finished in seconds. One normal credit pays for one normal search. Starter includes five a week, Daily Pro twenty-five a day and Business 2,500 a month; a free account starts with two credits in total. Credits do not roll over. Details are on the pricing page.

Deep search and deep credit

A deep search reaches further across the public web than a normal search and takes longer. It uses a deep credit instead of a normal one and needs an account. Every paid plan includes some deep credits; Daily Pro and Business can buy top-up packs. When it is worth running is covered on the deep search page.

Guest search

The free first search you can run on FaceSearch without creating an account. A short bot check runs first, then the search works like a normal search. Results from a guest search are not saved to a history because there is no account to save them to; sign in before searching if you want to come back to them later.

Source page

The public web page on which a matched image was found. The source page is the useful part of a result: it shows who published the photo, in what context, and gives you the address to send a removal request to. On the Free plan the top results are visible but the source links stay locked; paid plans unlock every link. See reading your results.

Thumbnail

The small copy of the matched image shown in a result list, so you can judge at a glance whether the face is yours before opening the source page. A thumbnail is a preview, not the original file: the full image stays on the site that published it, and removing a result from FaceSearch does not change that site.

What are the technology terms?

Face detection

The first step of any face search: finding where the faces are in a photo and drawing a box around each one. Detection does not know who anyone is; it only says a face is present. If detection finds no face, because the face is tiny, turned away or hidden, the search cannot continue. Photo tips are on the best photo for a face search.

Face embedding

A list of a few hundred numbers that describes the shape and features of one face, produced by a neural network from the detected face. Two photos of the same person produce embeddings that are close together; photos of different people produce embeddings that are far apart. Embeddings are what a face search compares. How they are made is explained on how face recognition works.

Similarity score and match score

A number describing how close two face embeddings are. A higher score means the faces look more alike to the model, which is why results are shown closest first. It is not a probability that the person is the same: a stranger with a similar face can score high, and a bad photo of you can score low. Treat a score as a ranking, not a verdict.

False positive

A result that scores high but shows a different person. False positives are unavoidable in any face search because many people share similar features and a web-scale index contains millions of faces. Confirm a match with non-face clues on the source page, such as the name, date, place or other photos, before acting on it.

Doppelgänger effect

The tendency of a face search to return look-alikes, relatives and twins alongside, or instead of, the person you searched for. It is a consequence of comparing similarity rather than identity, made more likely by low-quality photos and by the sheer size of the index. The doppelgänger effect page explains why it happens and how to work around it.

Index

The engine's stored collection of face embeddings, each tied to the public page it came from. When you search, your photo's embedding is compared against the index, not against the live web. A page that has never been crawled is not in the index and cannot appear in results, which is one reason a missing result proves nothing.

Crawl

The process of visiting public web pages, finding the images on them and adding any detected faces to the index. Crawling is continuous and never complete: new pages appear faster than any crawler can visit them, and pages that are removed or changed drop out over time. This is why results for the same photo change between searches.

Public web

Pages anyone can open without logging in: public profiles, news sites, blogs, forums, galleries, company pages. FaceSearch only indexes the public web. Private profiles, stories, direct messages, members-only sites and anything behind a password are outside it, by design. The platform pages say exactly where that line falls on each app.

What are the privacy and removal terms?

EXIF metadata

Hidden information a camera or phone stores inside a photo file: the date, the device, camera settings and often the GPS location. FaceSearch strips this metadata from every upload before the search runs, so a photo you upload never carries your location into the search. How uploads are handled is on privacy by design.

Re-encoding

Decoding an uploaded image and saving it again as a fresh file. Re-encoding removes anything that was hidden in the original file, including metadata and any unexpected content, and produces a clean image for the search to work on. FaceSearch re-encodes every upload after checking that the file really is an image of the type it claims to be.

Opt-out

A request to a face search engine to stop showing your face in its results. An opt-out removes matches from that one engine's results; it does not delete the photos from the websites that host them, and each engine needs its own request. FaceSearch's opt-out is free and handled by email, as described on remove my photos; guides for other engines sit under opt-out guides.

DMCA

The Digital Millennium Copyright Act, the United States law under which the owner of a photo's copyright can ask a website or search engine to remove copies published without permission. It applies to the person who took the photo or owns its rights, not necessarily the person in it. The DMCA page explains what a notice needs and what FaceSearch can and cannot remove.

Consent

Permission, given freely and knowingly, for a specific use of a photo or of the biometric data taken from it. Consent is the basis on which you search your own face, and the thing a stolen photo or a fake profile lacks. Under GDPR and similar laws, processing biometric data to identify someone generally needs explicit consent or another narrow legal basis.

What are the safety and scam terms?

Catfish

Someone who runs a dating or social profile with another person's photos and an invented identity. Because the photos are real pictures of a real person, they usually exist elsewhere on the public web, which is what a catfish check looks for. A clean result does not prove a profile is genuine; new and AI-generated photos leave no trail.

Romance scam

A fraud in which someone builds an online relationship over weeks or months and then asks for money, typically with a story that explains why meeting is impossible. The profile photos are often stolen from service members, doctors or models. The romance scam photo check page covers the photo search, the non-photo checks and where to report.

Sextortion

Blackmail with intimate images: a threat to share real or faked images unless the victim pays or sends more. Victims are told not to pay, to keep the evidence and to report to the platform and the police. Help routes come before any search tool on the intimate image abuse page.

Impersonation

An account that pretends to be you, usually built from your public photos and name. Every major platform has a report route for it, and a face search can find every copy of your photos before you file. Steps are on fake profile impersonation and how to report a fake profile.

NCII

Non-consensual intimate imagery: sexual or intimate images shared without the consent of the person shown, whether real, edited or AI-generated. Services such as StopNCII.org let you hash an image on your own device so participating platforms can block it, and Google and most platforms have dedicated removal forms. Help lines and routes are listed on intimate image abuse.

Deepfake

A synthetic image or video in which a face has been generated or swapped by software. A deepfake of you has no original photo to find, so a face search cannot detect it directly; it can only find public pages that reuse your real photos or that publish the fake. What a search can and cannot do here is on deepfake and AI image check.

OSINT

Open-source intelligence: research done with publicly available information, used by journalists, fact-checkers and investigators. A face search is one OSINT tool among many, and its result is always a lead to corroborate rather than a finding. Rules for professional use are on the pages for journalists and researchers and investigators.

What are the legal terms?

Biometric data

Data derived from a person's physical characteristics that can identify them, such as a face embedding, a fingerprint or an iris scan. Most privacy laws treat it as a special category with stricter rules. Whether a face search engine processes biometric data lawfully depends on the country, the purpose and whose face it is; see is face search legal.

BIPA

The Illinois Biometric Information Privacy Act, the strictest biometric law in the United States. It requires informed written consent before collecting a scan of face geometry and lets individuals sue for violations, which is why many face recognition products exclude Illinois. The United States page of the legal guide explains it alongside the Texas and Washington laws.

GDPR

The European Union's General Data Protection Regulation. Article 9 classes biometric data used to identify a person as special category data, which may only be processed on narrow grounds such as explicit consent, and Article 17 gives the right to erasure that opt-outs rely on. The EU page covers it and the regulators' actions against face search engines.

DPDP Act

India's Digital Personal Data Protection Act 2023, the country's first comprehensive data protection law, which sets consent and purpose rules for personal data and is being brought into force in stages. What it means for face search in India, together with the IT Rules and the national cybercrime reporting portal, is on the India page.

How does this glossary apply to FaceSearch?

FaceSearch is a reverse face search engine over the public web: it detects the face in your photo, computes an embedding, compares it with the index and returns source pages ranked by similarity. It never claims an identity, it strips metadata and re-encodes every upload, and it offers a free opt-out. Every term above is used on this site in exactly the sense given here, and the responsible use page sets out which uses of these tools are allowed.