Face Recognition: How It Works, Where It’s Used and the Debate

Face recognition (also called facial recognition technology) is the technology that knows your face. It unlocks your phone and checks your passport at the border. In some places it scans crowds for wanted people, which makes it one of the most debated technologies of our time.

This page explains what face recognition is and how it works. It also gives a quick tour of the wider topic with a link to our full guide on each part.

What is face recognition?

Face recognition is software that identifies or confirms a person from an image of their face. It measures the shape and features of a face and compares them with faces it already knows.

How does face recognition work?

A camera captures a face and software turns its features into a template made of numbers. It is compared with one stored face to confirm who you are or with many faces to find out who you might be. A match means the similarity score passed a set threshold and is never a certainty.

Why does face recognition matter?

Face recognition is now part of phones and airports as well as banks and policing. A face can be scanned from a distance without consent, which raises bigger privacy questions than most other biometrics. In many countries courts and lawmakers are now deciding how it may be used.

How Face Recognition Works, Step by Step

Every face recognition system follows the same basic pipeline. The camera and the software change from system to system. The steps stay the same.

Every face recognition system runs the same five steps1. DetectFind each facein the image2. AlignTurn and scale itto a standard pose3. EncodeTurn the face intoa template of numbers4. CompareScore it againstone or many templates5. DecideMatch only if thescore beats a thresholdA liveness check usually runs alongside step 1 to reject photos, screens and masks.Diagram: Biometric Today.
How face recognition works, step by step.

Step 1: Face detection

First the software has to find a face. Face detection scans an image and draws a box around anything that looks like a face. It answers “is there a face here?” and nothing more.

Detection runs on almost every phone camera today. It’s the reason your camera app can focus on faces. On its own it does not identify anyone.

Step 2: Alignment

Faces in photos are tilted and lit in different ways. So the system finds landmarks such as the eyes and the corners of the mouth. It then rotates and scales the face into a standard position.

Step 3: Feature extraction

Next comes the clever part. A neural network studies the aligned face and turns it into a template. That’s a list of numbers that describes the face in a way the computer can compare.

So what does face recognition look for? Modern systems don’t measure a fixed list of distances by hand. The network learns which patterns separate one face from another, from the shape of the eyes to the texture of the skin.

Step 4: Comparison

The new template is compared with stored ones. Two templates from the same person sit close together while templates from different people sit further apart. The system turns that distance into a similarity score.

Step 5: The decision

The score is checked against a threshold. Above it the system reports a match. Below it there is no match.

Where that line sits is a trade-off. A strict threshold rarely accepts the wrong person but rejects the right one more often. A loose threshold does the opposite.

Liveness detection

Good systems also check that a real live person is in front of the camera. This is called liveness detection or presentation attack detection. It tries to reject printed photos and masks as well as videos played on a screen.

The same steps power the biometric authentication on your phone and the cameras at an airport gate. What changes is the camera quality and how many faces the system compares against.

Two Jobs: Verification and Identification

Most face recognition does one of two jobs. Verification asks whether you are who you claim to be. Face ID does this and so does a passport gate comparing you with your passport photo.

Identification asks who a face belongs to by searching a watchlist or a database. Police searches, shop watchlists and face search engines all work this way. That’s where most of the controversy sits.

Verification checks one face against one record; identification searches manyVerification (1:1)Your passportor phone record“Are you who you say you are?”Face ID, passport e-gates, banking appsIdentification (1:N)“Who is this?”Police searches, shop watchlists, face search enginesMost of the debate around face recognition is about identification.
Face verification vs face identification.
Biometric authentication

Read the full guide

Biometric Authentication Explained: Methods, Steps and Trade-offs

How one-to-one face checks fit into logins and payments.

2D, 3D and Live Cameras: Not All Systems Work Alike

Not all face recognition works the same way. Most systems use ordinary 2D photos or video. That is cheap but easier to fool with a printed picture.

3D systems also measure the depth of the face. Apple’s Face ID projects more than 30,000 invisible dots to map it (Apple). Infrared cameras can see faces in the dark while live facial recognition scans every face in a video stream against a watchlist.

Face recognition apps on a phone

Read the full guide

The Face Recognition App on Your Phone: What It Does and What It Knows

How phone face unlock differs between brands and what photo apps know about you.

From a 1960s Lab to Your Pocket

  • 1964: Woody Bledsoe makes the first computer attempt at face recognition (arXiv).
  • 2000: NIST runs its first Face Recognition Vendor Test, still the benchmark today.
  • 2014: Facebook’s DeepFace shows that deep learning can match faces with 97.35% accuracy on a standard test.
  • 2017: Apple launches Face ID and puts 3D face recognition into millions of pockets.
  • 2020s: Face checks spread to airports, banks and shops as lawmakers start to set limits.

Can a Machine Get Your Face Wrong?

The best systems are very accurate on good photos. In NIST’s 2026 tests the top algorithms missed only about 1 in 1,000 searches of a mugshot database. With photos from self-service kiosks that rose to about 5% (Biometric Update).

Masks and poor light both hurt accuracy (NIST). Some algorithms have also made more errors for certain demographic groups. The most accurate ones were the most even (NIST).

On phones the odds are published by the makers. The chart below compares Face ID with fingerprint unlock.

A stranger is 20 times less likely to unlock Face ID than Touch IDOdds that a random person unlocks the device, as stated by Apple and Google1 in 1,000,000Face IDiPhone, 3D depth map1 in 50,000Touch IDone enrolled finger1 in 50,000Android floorany face or fingerprint unlockLonger bar = harder for a stranger to get in. The Android figure is the minimum Google allows, not a typical result.
Face ID vs fingerprint: odds that a random person unlocks the device.

Where You Meet It in Daily Life

Face unlock on phones is the most common use. Banks and apps ask for a selfie to open accounts while some shops scan customers against watchlists. Police use it to search mugshot databases for leads.

Airports are growing fast. In IATA’s 2026 survey 58% of travelers had used biometrics on an airport journey in the past year (IATA).

North America leads biometric use at airports, Europe trailsShare of travelers who used biometrics on their airport journey in the past 12 months, 2026North America63%Asia-Pacific61%Middle East61%Global average58%Latin America & Caribbean51%Africa49%Europe47%Source: IATA Global Passenger Survey 2026 (over 10,000 responses from more than 200 countries).
Biometrics at airports by region, IATA Global Passenger Survey 2026.
Face recognition applications

Read the full guide

Top 10 Face Recognition Applications with Live Examples

Real examples of face recognition at work around the world.

Airport facial recognition

Read the full guide

21 Airport Facial Recognition Adoption Examples

How airports use face scans from check-in to boarding.

Your Face or Your Fingerprint?

Face recognition needs no contact and works from a distance. That makes it convenient and it’s also why it raises more privacy questions than a fingerprint. The chart sums up the main differences.

Face vs fingerprint: same job, very different trade-offsFace recognitionFingerprint1. How it reads youShape of your face, via a camera2D photo or 3D depth mapRidges and valleys on a fingertipOptical, capacitive or ultrasonic2. AccuracyBest on 3D systems like Face IDDrops with masks and poor lightSteady on clean, healthy fingersStruggles with wet or worn skin3. Spoofing2D unlock can fall for a photo3D depth sensing is much harder3D-printed fakes worked in testsTakes a good print and real effort4. ConvenienceHands-free, works at a glanceAwkward lying down or in a maskFast and works in the darkNeeds a free, dry finger5. Who it works forWeaker for twins and under-13sError rates vary by demographicHarder for adults over 65Ridges wear with age and labour6. Privacy and consentCan be captured from a distanceYour face is already onlineYou have to touch a sensorHard to collect without you knowing7. Where it shows upPhones, airports, police searchesCan scan a whole crowdPhones, laptops, borders, locksOne person, one touch at a time
Face recognition vs fingerprint: the 7 differences at a glance.
Face recognition vs fingerprint

Read the full guide

Facial Recognition vs Fingerprint: Convenience, Security and the Trade-offs

Seven differences that matter from speed to privacy.

Why Face Recognition Is So Debated

Supporters say face recognition helps find missing people and stop identity fraud. It also makes phones and airports quicker to get through.

Critics point to wrongful matches. In 2020 Detroit police arrested Robert Williams after a false face match. The city later paid him $300,000 and tightened its rules (Michigan Public).

Scraping is another worry. Clearview AI was fined €30.5 million in the Netherlands for building a face database from photos copied off the web (DutchNews). The chart shows which public face search tools match the same face.

Only a few tools match the same face, and one is for police onlyMatches the same image or similar-looking photosMatches the same faceWhat it matchesAnyoneRestrictedImage lookup for everyoneTinEyeGoogle LensBingYandex ImagesYandex is the strongest of these on facesFace search for everyonePimEyesFaceCheck.IDLenso.aiIndex public web pages; most offer removalFace search for police onlyClearview AIBarred from most US private sales since 2022Restricted image lookupNot covered here
8 facial recognition search engines, sorted by what they match and who can use them.
Face search engines

Read the full guide

Face Search Engines: What They Find, What They Miss and How to Opt Out

How tools like PimEyes work and how to remove your face.

CCTV face recognition

Read the full guide

How CCTV Face Recognition Works in Surveillance

Where cameras recognize faces in public spaces.

Who Makes the Rules Around the World?

The rules differ a lot from one country to the next. Here is a quick overview.

RegionMain rules
European UnionThe AI Act bans scraping faces to build databases and mostly bans live face scanning in public by police.
United KingdomNo single face recognition law; data protection law applies.
United StatesNo federal law. Illinois requires consent, and San Francisco banned police use in 2019.
ChinaNational rules since June 2025 limit face recognition where other methods work.
IndiaThe Digital Personal Data Protection Act 2023 covers face data.
BrazilThe LGPD applies, and mobile lines need face checks from 2027.

Sources: eucrim, Hunton and Al Jazeera.

Biometric data

Read the full guide

What Is Biometric Data? Examples, Templates and What the Law Counts

Why faces count as sensitive data and what the law says.

Where Face Recognition Goes From Here

Deepfakes are the newest threat. Gartner said injection attacks rose 200% in 2023. In these attacks fake video is fed straight into an app (Gartner).

At the same time face checks are spreading into new jobs. Apple now uses a selfie to estimate age on the iPhone while airlines want one face scan to replace boarding passes.

Future of biometrics

Read the full guide

Your Body Is the New Password: 8 Shifts Shaping the Future of Biometrics

Deepfakes and passkeys are among the shifts coming next.

Explore face recognition

Face recognition apps

From Face ID to Google Photos your phone uses face recognition every day. See what seven popular apps do and what they know about you.

Read the full guide

Face search engines

Tools like PimEyes can find your photos across the web from a single selfie. Learn how they work and how to opt out.

Read the full guide

Face vs fingerprint

Which is faster, harder to fool and kinder to your privacy? We compare the two most common ways to unlock a phone across seven differences.

Read the full guide

Airports and borders

Face scans now stand in for boarding passes and passport checks at many airports. See how airports around the world are using the technology.

Read the full guide

CCTV and surveillance

Cameras that recognize faces in public spaces are the most contested use of all. See how CCTV face recognition works and where it is used.

Read the full guide

Real-world applications

Beyond phones and passports face recognition shows up in banks, shops and hospitals. Explore real examples of how it is being applied.

Read the full guide

2D, 3D and live systems

2D and 3D systems work in different ways, and so do live cameras. Our upcoming guide compares them side by side.

Guide coming soon

Errors and false matches

How often does face recognition get it wrong and why? This guide will cover independent test results and what affects them.

Guide coming soon

Rules by country

Rules differ a lot from one country to the next. This guide will compare them.

Guide coming soon

Latest face recognition news

Face recognition guides and explainers

Face recognition by country

Country guides are on the way. Until then these stories show how different countries are handling face recognition:

References

Biometric Today
Logo