The PSL Scale, Explained: Where It Came From and What It Can't Measure

PSL is an abbreviation of the names of three online incel forums that no longer exist. Gemma Ware, writing in The Conversation in 2026, describes it plainly: 'The PSL scale is a pseudoscientific attractiveness rating system used by looksmaxxers.' The same article traces its roots to misogynistic forums used by incels (involuntary celibates) and notes that it has since moved onto mainstream social media, where teenage boys post photos of themselves asking to be ranked.

Sociologist Jordan Foster of MacEwan University, interviewed for that article, explains the appeal of dressing a ranking system up in numbers and jargon: 'Wrap it in the guise of science.' That framing is why the scale feels authoritative. It uses decimals, tiers and technical words like 'canthal tilt' and 'gonial angle', but none of that structure came from research.

Looksmaxxing has also reached the medical literature. In a 2026 commentary in Plastic and Reconstructive Surgery Global Open, Maini and Janis describe people comparing themselves with 'formulaic beauty ratios' such as the facial rule of thirds, the facial width-to-height ratio, positive canthal tilt and the gonial angle, and warn about partial understanding and misapplication of those measures.

There is no official version of the scale, because no organisation owns it. One of the top-ranking PSL sites, pslscale.com (read 2026-10-23), presents it as a 0–8 score split into seven bands, from a lowest band at 0–2 up to a top band at 7.75–8, with forum slang names for each band. We describe the structure here without the labels, since several of them exist to rank people's worth.

The most common question on forums is what percentile a PSL number represents, for instance whether a 5 is above average. The honest answer is that nobody knows, because nobody has measured it. There is no sample of rated faces behind the bands. pslscale.com itself says its ranges 'are not population statistics or a scientifically validated beauty standard' and that its score is 'not a population percentile'. Conversion tables you see posted online, between 1–10 'deciles' and PSL, are forum conventions, not data, which is why users argue about them.

A measurement scale earns trust in three ways: its scores are defined, they are checked against something real, and independent people get the same answer. The PSL scale has none of these. We found no peer-reviewed study that validates it, its tiers have no published definition, and the scores come from strangers on forums or from apps that don't publish their formulas.

The ratios PSL culture leans on don't hold up well either. The golden ratio, a favourite of rating apps, has been tested repeatedly. Kiekens and colleagues (2008) found that only 4 of 19 golden proportions correlated with ratings of 64 adolescents' faces, and together they explained 16% of the variance. Naini (2024) concluded there is 'no convincing evidence' linking the golden ratio to facial beauty. Khoshab and colleagues (2022) pooled raw data from 27 studies and found no ethnic or gender group that consistently matched either the classical canons or phi.

Automated raters add their own bias. Goshtasbi, Hakimi and Wong (2024) compared five AI rating websites with 24 human raters on 40 faces. The AI sites' scores tracked the human averages (r = 0.84) but ran high: 6.9 on average against 5.0 from people, with 30 of the 40 faces scored significantly higher by the sites. A number that looks precise can be off in a consistent direction, and you would never know from the screen.

Real research on faces isn't empty, it just says something different from PSL. People do agree about faces when you average many raters. A 2000 meta-analysis by Langlois and colleagues found that the averaged ratings of adult raters were highly reliable within a culture (r = .90), across ethnic groups (r = .88) and across cultures (r = .94). Those figures describe the average of a group of raters, not how much two individuals agree.

Individual taste is large. Hönekopp (2006) split attractiveness judgements into shared and private taste and found that 'private taste is about as powerful as shared taste'. In other words, roughly half of the meaningful variation in how one person rates a face is personal to that person. A single number cannot describe a face for everyone.

Some features do matter on average. Rhodes (2006) reviewed the evidence that averageness, symmetry and sex-typical features are each found attractive in men's and women's faces and across cultures. Langlois and Roggman (1990) showed that composite faces, made by averaging many photos, were judged more attractive than almost all of the individual faces in them. But the effects are modest: Jones and Jaeger (2019) note that meta-analytic estimates put symmetry's effect at about r = 0.25, the smallest of the three, and found that these factors explain less variance in natural faces than the literature suggests.

Photos matter more than people think. Bryan, Perona and Adolphs (2012) photographed the same faces at 45 cm and 135 cm; the close-up versions were rated less attractive, trustworthy and competent. A rank given to a selfie partly measures the camera.

FaceCheckup's Balance Score runs from 0 to 10 and is built only from the two things above that can be measured on a photo with a published reference. The symmetry part (up to 6 points) mirrors 220 pairs of landmarks across the centre of your face and averages how far each point lands from its partner, as a percentage of face width. The first 0.8% is free, then the score drops by 1 point for every further 0.6%.

The proportion part (up to 4 points) uses only the two ratios Pallett, Link and Lee (2010) tested: distance between the pupils at about 46% of face width, and eyes-to-mouth distance at about 36% of face length. Within 2 percentage points of either value costs nothing; beyond that, each ratio loses 1 point per 2 points, up to 2 points each. Pallett's experiments used female faces and measured width between the ears, so we say so next to the result.

Before it measures, the scan rotates your head pose back to frontal, because a slightly turned head looks asymmetric. In our own test on one selfie with the head turned about 10°, raw asymmetry read 4.70% of face width and 1.06% after pose correction. That gap is why we refuse to score bad photos and show every number with a ± range. The full formula is on the how-it-works page, and the facial symmetry test page explains the mirror method in detail.

It would be easy to map a 0–10 Balance Score onto a 0–8 PSL scale, and some competitors do something similar. We don't, for four reasons.

The Conversation's reporting notes that if you have teenagers in your life, they have probably heard of the PSL scale, and that looksmaxxers sometimes use extreme methods to change their appearance. Grillo and colleagues (2026) tracked practices such as 'bonesmashing' on social media and described them as dangerous trends being normalised in online communities. We give no advice on exercises, fillers, surgery or other procedures that claim to change the face.

Checking your face over and over, comparing it with strangers' ratings, and feeling that a flaw others barely notice is ruining your life can be a sign of something more than a bad day. The NHS describes body dysmorphic disorder (BDD) as a mental health condition in which a person spends a lot of time worrying about flaws in their appearance that are often unnoticeable to others. It is treatable. The NHS advises seeing a GP if you think you might have it, and notes that symptoms probably won't go away without treatment. If you are under 18, a family doctor can refer you to young people's mental health services; wherever you live, a doctor or school counsellor is a good first step.

If you came here wanting to look better in pictures, the evidence points to the camera and the grooming choices you control rather than to a rank. These come from FaceCheckup's sourced rulebook, with their labels.