Your face shape is one of seven categories most stylists and analysts use: oval, round, square, heart, oblong (rectangle), diamond, or triangle (pear). You figure it out by comparing four measurements (face length, forehead width, cheekbone width, and jaw width), then matching the ratios to a category.
That is the short version. The longer version is more interesting, because the categories themselves are fuzzier than the internet pretends, the "golden ratio" stuff is mostly junk, and AI is genuinely useful here for one specific reason: it does not care what you want the answer to be.
Face shape categorization comes out of cephalometry, the orthodontic and anthropological practice of measuring head and face proportions. Clinical research uses a measure called the facial index (face height divided by face width) to sort faces into types like mesoprosopic (medium), leptoprosopic (long), and euryprosopic (broad). The styling categories you see on Pinterest are a simplified version of that, with the jawline and forehead width added in.
Here is how each shape breaks down:
The "default" reference shape in most styling literature. Face length is roughly 1.5× the width of the cheekbones. Forehead is slightly wider than the jaw, and the jawline is rounded rather than angular. Oval is often described as the most "balanced" shape, mostly because it sits closest to the population average for facial proportions.
Celebrity examples often cited: Beyoncé, Bella Hadid, Jessica Alba.
Face length and cheekbone width are roughly equal. The jawline is soft and curved with no hard angles, and the widest point is the cheeks. Round faces tend to read younger because they share proportions with infant faces, a feature linked to perceived youth across cultures in facial averageness research by Rhodes and colleagues.
Examples: Selena Gomez, Chrissy Teigen, Adele.
Face length and width are again roughly equal, but the jawline is angular and the forehead, cheekbones, and jaw are all about the same width. The defining feature is the angle at the jaw, not the overall ratio. Think of a square face as a round face with corners.
Examples: Olivia Wilde, Angelina Jolie, Demi Moore.
Forehead and cheekbones are wide; the face tapers to a narrow, often pointed chin. Sometimes called an "inverted triangle." If you have a widow's peak, you are statistically more likely to read as heart-shaped, though hairline pattern alone does not determine face shape.
Examples: Reese Witherspoon, Scarlett Johansson, Kourtney Kardashian.
A longer version of the square. Face length is noticeably greater than width (often 1.6× or more), with a straight jawline and forehead, cheekbones, and jaw of roughly equal width. In clinical anthropometry this corresponds to the leptoprosopic type, which one cross-sectional study of dental students found was the single most prevalent face type in its sample at around 40%.
Examples: Sarah Jessica Parker, Liv Tyler.
Cheekbones are the widest point; forehead and jawline are both narrower, and the chin is pointed. Diamond is rarer than the others in most population samples and is sometimes mistaken for heart-shaped. The difference is that heart-shaped faces are widest at the forehead, while diamond faces are widest at the cheeks.
Examples: Halle Berry, Jennifer Lopez (debated; she also frequently gets classified as oval).
The inverse of heart. Jawline is the widest point; forehead is the narrowest. Less common than the other six in most reference sets, partly because it is often grouped with "square" or "round" by less careful analyses.
Examples: Minnie Driver, Kelly Osbourne.
A note before you keep reading: most faces are not pure examples of one category. The honest answer for a lot of people is "oval-leaning-round" or "square-leaning-oblong." Don't force yourself into one box.
You need a tape measure (a soft fabric one if you have it), a mirror, and a hair tie. Pull your hair back so you can see your full hairline and jawline.
Take these four measurements:
Now compare:
If two categories feel equally true, you are probably between them. That is normal. Face shape is a continuous distribution that we have arbitrarily chopped into seven bins for styling convenience.
AI face shape detection works in three steps. First, the model runs facial landmark detection, identifying a set of reference points on your face (eye corners, nose tip, jawline contour, hairline). Modern landmark detection networks predict 68 to 468 of these points per face. According to a single-shot detection model published in PMC, well-tuned models can hit around 99% face detection precision with an average landmark error of roughly 2.3 pixels.
Second, the model extracts the same ratios you would measure with a tape measure: length-to-width, forehead-to-jaw, cheekbone position. Third, it matches those ratios against learned examples of each shape category.
The whole pipeline is mature enough that the technical part is rarely the failure mode. The failure modes are about the photo:
The other thing worth flagging: most public face shape models were trained on datasets that skew toward Caucasian features. A 2018 review in Frontiers in Genetics on facial genetics notes that facial morphology varies significantly across ancestry groups in ways that simple seven-bin categorization does not capture cleanly. If you have East Asian, African, South Asian, or mixed ancestry, expect a bit more category drift between tools.
Want the AI read for your specific photo? Our Face Shape Analyzer runs the full landmark-and-ratio analysis on a selfie and returns a category with confidence. Free, no signup, instant. Pair it with the Facial Harmony and Jawline Analyzer tools if you want more depth on individual features.
The honest answer: not much. Or rather, not what TikTok and Pinterest claim.
Face shape correlates with some things and not with others. Here is what the research actually supports:
Real, with evidence: Face shape affects what hairstyles and frames look balanced on you. This is a geometry observation, not magic. It also correlates with perceived age: rounder, fuller faces tend to read younger because they share proportions with juvenile faces, a finding consistent with Rhodes et al.'s work on perceived health and attractiveness.
Real, but smaller than people think: Attractiveness research finds that averageness (being close to the population mean for facial proportions) has a modest but consistent effect on attractiveness ratings across cultures. Rhodes and Tremewan (1996) showed that moving facial features away from the average reduces attractiveness ratings. This is part of why oval (close to average) shows up as a "safe" category, not because it is objectively prettier, but because it is statistically more average.
Mostly bunk: The idea that your face shape predicts your personality, intelligence, or fate. This is physiognomy, it has been repeatedly debunked, and modern facial-analysis studies that find tiny correlations between bone structure and behavior are working with effect sizes that disappear when you control for confounds.
Mostly bunk, version 2: The Marquardt "Phi Mask," the idea that the golden ratio (≈1.618) is the gold standard for facial beauty. The mask was popularized by a 2006 study by Kar Bashour using 224 reference points, but a follow-up critique published in Aesthetic Plastic Surgery (Holland (2008), "Marquardt's Phi Mask: Pitfalls of Relying on Fashion Models and the Golden Ratio") pointed out that the mask was built from a sample of Western fashion models and does not generalize across ethnicities or even across non-model Caucasian faces. The "ideal face" the mask depicts is closer to a masculinized Northern European fashion archetype than to any biological optimum. Treat phi-based attractiveness scores as a fun aesthetic exercise, not a verdict.
If you want the longer version of why the golden ratio gets wildly overclaimed, we wrote a whole separate piece on it.
The geometry rule of thumb most stylists use: contrast your face shape, do not match it. Angular features look softer with round shapes; round features look more defined with angular shapes. Here is the short version:
| Face shape | Glasses that work | Hair that works | Makeup that works |
|---|---|---|---|
| Oval | Almost anything; geometric frames add interest | Most cuts work; layered medium lengths are flattering | Light contouring; no major correction needed |
| Round | Rectangular or angular frames (wider than tall) | Layers, height at the crown, long side parts | Contour under cheekbones to add definition |
| Square | Round or oval frames; soft curves | Soft waves, layers around the jaw | Soften jaw corners with blush placement |
| Heart | Bottom-heavy frames (rimless tops, bold bottoms) | Chin-length cuts; side-swept bangs | Highlight chin; soften forehead with bronzer at temples |
| Oblong | Tall frames or oversized round shapes (break length) | Bangs, curls, anything adding horizontal volume | Horizontal blush placement to shorten visual length |
| Diamond | Oval or cat-eye frames; emphasize forehead | Side-swept bangs; chin-length cuts | Soften cheekbones; highlight forehead and chin |
| Triangle | Top-heavy frames; embellished or bold tops | Volume at the crown, shorter sides | Highlight forehead, contour jawline |
These rules come from optical retailers and styling literature (Warby Parker's frame guide and All About Vision both cover the geometry in detail). They are aesthetic conventions, not laws. Plenty of people break them and look great. The point of the rules is to give you a starting frame of reference, not a cage.
A handful of claims that show up constantly and are mostly wrong:
"You can change your face shape with mewing." No credible evidence. The American Association of Orthodontists has stated there is no current research showing tongue posture changes adult jawline structure. Most "mewing before-and-after" photos are explained by posture, body fat loss, and camera angle.
"The golden ratio is the universal standard for facial beauty." Covered above. The Marquardt mask is a fashion-model-derived template and does not generalize across populations.
"Face shape determines personality." That is physiognomy, debunked for over a century. Modern studies claiming small correlations rarely survive replication.
"Oval is objectively the most attractive." Oval reads as attractive partly because it is closest to the population average, and averageness modestly predicts attractiveness ratings. That is a statistical effect, not a verdict.
"Your face shape is fixed forever." Mostly true. Weight changes, aging (which thins facial fat pads), and dental or surgical work shift apparent shape. Underlying bone structure is stable in adulthood.
A few honest takes if you got this far:
Want the AI read on your own photo? Start with the Face Shape Analyzer, then dig into the Facial Harmony and Jawline Analyzer tools for individual features. All free, no signup.