Looksmaxxing & face-analysis myths, fact-checked

Symmetry is one signal of attractiveness, not the dominant one. Averageness — how close a face is to the population mean for its sex and ethnic group — explains more variance in perceived attractiveness than symmetry alone in controlled studies.

Studies measuring fluctuating asymmetry against rated attractiveness consistently find a small-to-moderate effect, but averageness, sexual dimorphism, and skin quality each carry comparable or greater weight. Rhodes (2006) reviewed three decades of research and concluded symmetry is one of multiple cues, not a master signal. Most natural human faces sit within a narrow asymmetry band; the differences that matter are sub-millimeter and often invisible to the naked eye. A face report that scores symmetry alone is missing the larger half of the picture.

The golden ratio is a useful proportion heuristic — it correlates loosely with attractiveness ratings — but it is not deterministic. Faces that score lower on phi-based metrics are routinely rated more attractive than faces with closer-to-1.618 proportions.

Marquardt (2002) popularized the phi-mask, but later peer-reviewed work has not replicated a strong predictive relationship. Holland (2008) and Pallett, Link & Lee (2010) found the new "golden ratio" of facial proportions for attractiveness was empirically closer to 1.45 (vertical) and 0.46 (horizontal), not 1.618. The phi value works as a directional signal, not a rule. Tools that grade attractiveness solely on golden-ratio adherence — including some popular face-rating sites — are reporting on one weakly-predictive variable.

No peer-reviewed study has demonstrated that adult mewing produces measurable bony changes to the maxilla or mandible. Soft-tissue and posture differences are real, but the dramatic before/after photos circulating on social media are typically lighting, angle, weight, and bodyfat changes — not bone remodeling.

Bone remodeling in adulthood requires sustained orthopedic load — orthodontic appliances, surgical advancement, or bite-block expansion — not tongue position. Adolescent mewing during the growth window has weaker but plausible support; adult claims do not. The American Association of Orthodontists has issued no endorsement of mewing as an alternative to orthodontic treatment. Posture-driven jaw clarity is real and worth pursuing, but anyone marketing mewing as a substitute for ortho or surgery is overselling the evidence.

FWHR correlates weakly with dominance perceptions and some behavioral measures in men, but the effect is small and heavily debated. Recent meta-analyses find the association is weaker than early papers suggested and may not replicate across ethnic groups.

Carré & McCormick (2008) reported FWHR predicted aggression in hockey players. Subsequent replications produced mixed results. Kosinski (2017) and Geniole et al. (2015) ran larger samples and found the effect was either small or attenuated after controlling for body size. FWHR is real and measurable — it is one of the metrics in the RealSmile face report — but treating it as a personality predictor is overreach. Treat it as a perception cue, not a destiny.

AI scoring is consistent — it returns the same number for the same photo every time — but consistency is not objectivity. The ratings every model is trained on come from human rater panels, and those panels carry the cultural, ethnic, and demographic biases of their composition.

Every face-rating model is a function of its training data. If the rater panel skews young, urban, and Western, the model encodes those preferences. If photos in the training set were lit from above with neutral expressions, the model will score those same conditions higher in production. The RealSmile face report is calibrated against published landmark research and discloses its rater composition; we do not claim objectivity, we claim transparency about the inputs.

Angle, lens focal length, lighting, and camera height each move composite face scores by 5–15 points on a 100-point scale. The same person photographed in two conditions will get materially different ratings from any landmark-based scoring engine.

A 24mm lens at chin height distorts the lower face and inflates jaw width measurements; a 50–85mm lens at pupil height reads geometry as the eye sees it. Top-down lighting deepens orbital shadows and reads as more masculine; flat front lighting reads softer. Tilt of even 5 degrees off-axis changes apparent canthal tilt, midface ratio, and FWHR. The first hundred words of any face-report methodology should warn the user about photo-condition variance — if it does not, treat the score with skepticism.

Halo-effect research consistently shows that perceived attractiveness correlates with hiring decisions, pay, dating-app match rates, and social-judgment outcomes. Whether looksmaxxing is "worth it" is a values question, but the upstream perception effect is well-documented.

Hamermesh & Biddle (1994) found a 5–10% wage penalty for below-average-looking workers and a 3–5% premium for above-average. Mobius & Rosenblat (2006) showed the wage effect persists after controlling for productivity. On dating platforms, photo audit data shows roughly 60% of swipe outcomes are determined by the lead photo. The honest framing: lookism is real, looksmaxxing is one response, and the ethical questions about responding to it are separate from whether the underlying effect exists.

Bone structure is fixed without surgery, but apparent face shape is heavily moderated by body fat, hydration, lymphatic state, posture, and skin elasticity. Visible changes within an 8–12 week window are common and produce real differences in face-report scores.

Buccal fat distribution, masseter hypertrophy or atrophy, supraorbital fat pad volume, and nasolabial fold depth all shift with bodyfat percentage and hydration. A leaner, well-rested, hydrated face reads as more angular even with identical underlying bone. Skin-quality interventions (SPF discipline, retinoid use, hydration protocols) compound over years. The static "face shape is destiny" framing common in looksmaxxing forums is incomplete — the dynamic component is small but cumulative.