The FaceTest Blog
How AI face matching really works, how to take a photo that matches well, and where face-reading traditions like gwansang and animal face types come from — written by the team that builds the tests.
2027, the Year of the Red Goat: How to Read a Korean Zodiac Fortune
Why 2027 is jeongmi and a red goat year, when your sign changes, how the goat year relates to the twelve signs, which signs have samjae in 2027, and how the monthly flow is worked out.
How Can a Human Face Match a Pokémon? Shape Language, Explained
Creature designers encode personality in circles, triangles, and squares. That same shape language is what an AI reads in your face, which is why Pokémon matches feel oddly right.
Why Villains Have Those Faces: First Impressions, Typecasting, and What a Face Test Sees
We judge a face as trustworthy or dangerous in a tenth of a second, and casting directors know it. The research behind on-screen face types, and why your match says nothing about your character.
Glasses, Bangs, Beards, Makeup: How Much Does Styling Change Your Match?
A self-experiment guide: which styling changes move an AI look-alike result a lot, which barely matter, and how to separate your face's real signal from the costume.
Who Looks More Like Them? How to Run a Fair Face-Match Battle With Friends
Same test, same rules, honest scores: a short guide to settling the group-chat argument with battle mode, and reading the results without starting a new argument.
How AI Look-Alike Tests Actually Work
Face detection, embeddings, cosine similarity: a plain-language tour of what happens between uploading a photo and seeing your match, and why the analysis never leaves your browser.
How to Take a Photo That Actually Matches Well
Lighting, angle, expression, filters, framing: practical rules for a photo the AI can read, and why each one changes your result.
Why Your Friends Say You Look Like Someone (and You Don't See It)
Holistic face processing, the 'face space' idea, familiarity, and the mirror problem: what psychology says about resemblance, and how the AI's version compares.
Before You Upload Your Face: A 7-Point Privacy Checklist
Face data is biometric data. What to check in any face app's policy, what 'on-device' really means and how to verify it, and exactly how FaceTest handles your photo.
Why Anime Faces Are Hard for Face AI (and How We Match Them Anyway)
Giant eyes, missing noses, hair as identity: the domain gap between photos and animation, and the rules FaceTest follows to keep anime tests honest.
From Physiognomy to Face Filters: A Short History of the Look-Alike Test
Ancient face-reading, a 19th-century pseudo-science, celebrity lookalike contests, how-old.net, and the museum selfie: how we arrived at AI look-alike tests.
Gwansang: What Korean Face Reading Says, and What It Doesn't
The three zones, the five organs, the twelve palaces: a guide to Korea's face-reading tradition, its history and pop-culture revival, and how it differs from AI matching.
Dog Face or Cat Face? The Korean Animal Face Type Guide
Where 강아지상, 고양이상, 여우상 and the rest come from, which features define each type, and how an AI decides between them.
Do You Have a Doppelgänger? What Science Says About Look-Alikes
A 2022 genetics study of unrelated look-alikes, the 'one in 135' estimate, and why face-recognition algorithms are surprisingly good at finding your twin stranger.
What Does an 82% Similarity Score Actually Mean?
Cosine similarity in plain terms: why 100% never happens, why scores aren't comparable across tests, and what to look at instead of the number.
What's Your Face Shape? A Measure-It-Yourself Guide
Oval, round, square, heart, oblong, diamond: how to measure your own face in two minutes, and why face outline is one of the strongest signals in AI matching.
Symmetry, the Golden Ratio, and Other Face Myths
What research actually shows about symmetry, averageness, and the 'golden ratio' face, and why a look-alike AI doesn't care about any of it.
How a FaceTest Test Gets Made, From Pick to Publish
Choosing a work, collecting in-scene stills, one medium per test, twelve to sixteen characters, names checked in ten languages, and a QA gate: our editorial process step by step.