References · appearance

What actually shows up in the mirror: the evidence

Sleep, food and training change how a face reads to other people, and the change is measurable within weeks. 8 peer-reviewed sources, last verified 8 September 2026.

Read this before trusting the numbers

Every outcome in this document is a stranger rating a photograph. That is a real measurement and it is not a life. The samples skew young, light-skinned and at university, and the authors of the strongest paper here say so themselves. The effects are smaller than the topic's reputation. A night of lost sleep moved rated attractiveness by four percent, in twenty-three people. That study is quoted worldwide as proof of beauty sleep and it does not carry the weight. What it establishes is a direction and a mechanism: raters were reading tiredness. The randomized trial in here measured skin and stopped. AP-P-02 is the only randomized controlled trial in the document. It proves a daily smoothie changes the color of skin. It did not ask a single person whether the result looked better, and saying it did would be inventing an outcome nobody collected. One entry contradicts the rest and it is the reason this document exists in this form. AP-C-01 records that the level of facial fat rated most attractive is not the level rated healthiest. An app optimizing for how a face reads would push some members below their healthiest weight. MVIII sets targets from the member's stated goal and never from a scan, and this finding is one of the reasons that rule holds.

Grades: A multiple meta-analyses or systematic reviews in agreement. B one meta-analysis, or several consistent controlled trials. C limited or single trials, wide intervals, high heterogeneity, or cross-sectional, retrospective, biomechanical or survey designs only.

The entries

AP-P-01Sleep is programmed, and one lost night is visible to strangers

PrescriptionSleep is treated as programming, not as recovery advice, and a short night is a reason to change the day. Nothing in the app frames sleep as a way to look better. Band: min "do not schedule against sleep", standard "eight hours in the window the member keeps", max "measured sleep driving session load".

Evidence gradeC

Effect23 adults were photographed twice, once after at least eight hours of sleep and once after five hours in bed followed by 31 hours awake. 65 untrained observers rated the photographs on 100 mm scales. Sleep-deprived faces were rated less healthy (68 to 63), less attractive (40 to 38) and more tired (44 to 53), all at p < 0.001 [1]. With tiredness entered as a mediator the independent effect of sleep loss fell from -4.2 to -1.8 on health and from -1.6 to -0.62 on attractiveness.

Population23 photographed, aged 18 to 31, 12 women and 12 men, balanced and counterbalanced. 65 raters aged 18 to 61.

Sources[1]

CaveatFour percent, one acute night, twenty-three people. The mediation is the useful part: raters were reading tiredness, and tiredness is what moved. Nothing here tests habitual sleep over months, which is the thing anyone would actually want to know.

AP-P-02Vegetables are counted as food, and they change skin color

PrescriptionFruit and vegetable intake is programmed as food. This entry records that the effect also shows up in a mirror, which is not a reason the app gives for it. Band: min "some fruit or vegetables daily", standard "enough to shift carotenoid intake", max "measured carotenoid supplementation".

Evidence gradeB

Effect81 university students were randomized to a daily 500 mL fruit smoothie carrying about 25.4 mg of carotenoids, or to bottled water, for six weeks with a two-week follow-up. Skin yellowness rose 3.38 units by week five and held through follow-up (p < 0.001). Redness rose about 1.08 units (p < 0.001). Neither moved in the control group [2]. Both groups darkened slightly in luminance (1.73 units, p = 0.013), which the authors attribute to sun exposure and not to carotenoids.

Population81 Malaysian Chinese students, 34 men and 47 women, mean age 20.5. Completion 80.5% intervention, 85% control.

Sources[2]

CaveatThis trial measured color and nothing else. No rater saw these faces and no perception outcome was collected. The half of the claim about how the color reads comes from [3], a different design on a different sample. Both halves are needed and only one of them is randomized.

AP-P-03Conditioning and body composition, read off the skin

PrescriptionAerobic work and body composition are programmed for what they do to the body. This entry records a second route by which they show up. Band: min "any regular aerobic work", standard "the conditioning the program already prescribes", max "measured VO2 max targets".

Evidence gradeC

EffectHigher aerobic fitness and lower body fat each independently predicted greater skin yellowness. The cross-sectional model explained 49% of the variance in 134 people, and a longitudinal model explained 42% of the change over eight weeks in 59 people. A VO2 max difference of about 8.75 ml/kg/min produces a color change visible to the eye, and in the perception study participants chose the fitness-associated coloring on 85% of trials [3]. Skin lightness did not move, so the route is carotenoids and not melanin.

Population134 (mean age 21.3), 59 over eight weeks (mean age 21.1), 21 in the perception study.

Sources[3]

CaveatThe authors state the limits: restricted ethnic diversity, a narrow age range, predominantly light-skinned samples, and a note that carotenoid color may be harder to detect in darker skin. The eight-week arm is the strongest part and it is 59 students. A mechanism with a plausible size. Not a number to promise anyone.

AP-P-04Sun protection and skin care are given as skin advice

PrescriptionSun protection and skin care follow dermatology association guidance and are framed as skin health. Nothing in the app frames them as a way to be better looking. Band: min "daily sun protection", standard "the AAD basics", max "a dermatologist-led regimen".

Evidence gradeC

EffectFaces manipulated toward more homogeneous skin color were rated more attractive at levels where observers could not consciously report that anything had changed. Attractiveness perception followed a logarithmic function while physical change detection followed a linear one, and at 30% of the manipulation raters called faces more attractive while reporting no visible difference [4]. Earlier work established the direction in female [5] and male [6] faces and tied it to the distribution of melanin and hemoglobin.

Population36 undergraduates for change detection, then 21, 18 and 17 across three separate perception techniques [4].

Sources[4], [5], [6]

CaveatSmall samples throughout, and the manipulation is digital. No one applied a treatment to a face. The correlation magnitudes reported in [5] and [6] were not verified against the full text for this document and are cited for direction only. Nothing here shows that a skincare routine moves the measure, which is the step somebody would need before promising anything. # PART B — Findings

AP-C-01Where looking better and being healthier stop agreeing

FindingFacial fat predicts both rated health and rated attractiveness, and the two do not peak at the same level. Raters prefer a lower weight for attractiveness than for health [7].

Evidence gradeC, contested

EffectPerceived facial adiposity predicts rated health and attractiveness and tracks cardiovascular measures and reported infections [8]. Asked directly whether the most attractive level of facial adiposity is also the healthiest, the answer was no: the attractive optimum sits below the healthy one [7].

PopulationRating studies of adult faces, cross-sectional.

Sources[7], [8]

CaveatThis is the entry that keeps the rest of the document honest. Every other row here is a case where the healthy thing and the flattering thing are the same thing, and that overlap is what makes them safe to program against. This one is not. A feature moving a member toward the rated-attractive optimum would be moving them away from the rated-healthy one, on the strength of a cross-sectional rating study. MVIII does not do it.

What this program will not tell you

Things commonly prescribed with confidence that the research does not currently support. MVIII programs none of them.

  1. AP-X-01 — That a skincare routine measurably changes how attractive a face is rated. Skin color homogeneity is associated with the ratings [4], [5], [6], and every one of those manipulations was digital. No trial located tested a skincare regimen against a rated-appearance outcome. The app gives sun protection and skin care on dermatology guidance, for skin, and claims nothing further. The gap between "evener skin rates better" and "this routine will make you rate better" is two studies wide and nobody has run them.
  2. AP-X-02 — That facial exercises reshape a face. Facial exercise has one trial behind it, on perceived facial age in middle-aged women, and it is the citation the app already shows on the facial scan. It is not evidence for jaw training, tongue posture, or any of the structural claims made online. Bone does not remodel because somebody clenched. MVIII returns facial findings as an ordered list of changes and rates nothing, which is the position this row supports.
  3. AP-X-03 — That there is a number for how attractive someone is, and that an app should compute it. Rating studies produce group averages under laboratory conditions. They do not license handing an individual a score. MVIII does not score a face, and SI-H-01 in the self-image document forbids the app from making a claim about a member's body in either direction. This row exists so the refusal is recorded with the rest of the evidence rather than only in the code.
  4. AP-X-04 — That optimizing appearance and optimizing health are the same project. They overlap in every entry in Part A, and AP-C-01 is where they part. Treating them as identical is the assumption that turns an appearance feature into a weight-loss instruction the member never asked for.

References

  1. Axelsson J, Sundelin T, Ingre M, Van Someren EJW, Olsson A, Lekander M. Beauty sleep: experimental study on the perceived health and attractiveness of sleep deprived people. BMJ. 2010. https://doi.org/10.1136/bmj.c6614
  2. Tan KW, Graf BA, Mitra SR, Stephen ID. Daily Consumption of a Fruit and Vegetable Smoothie Alters Facial Skin Color. PLOS One. 2015. https://doi.org/10.1371/journal.pone.0133445
  3. Foo YZ, Rhodes G, Simmons LW, et al. Skin Color Cues to Human Health: Carotenoids, Aerobic Fitness, and Body Fat. Frontiers in Psychology. 2020. https://doi.org/10.3389/fpsyg.2020.00392
  4. Your face looks the same as before, only prettier: The facial skin homogeneity effects on face change detection and facial attractiveness perception. Frontiers in Psychology. 2022. https://doi.org/10.3389/fpsyg.2022.935347
  5. Matts PJ, Fink B, Grammer K, Burquest M. Color homogeneity and visual perception of age, health, and attractiveness of female facial skin. J Am Acad Dermatol. 2007. https://doi.org/10.1016/j.jaad.2007.06.040
  6. Fink B, Matts PJ, Brauckmann C, Gundlach S. Colour homogeneity and visual perception of age, health and attractiveness of male facial skin. J Eur Acad Dermatol Venereol. 2012. https://doi.org/10.1111/j.1468-3083.2011.04316.x
  7. Coetzee V, Re D, Perrett DI, Tiddeman BP, Xiao D. Judging the health and attractiveness of female faces: Is the most attractive level of facial adiposity also considered the healthiest? Body Image. 2011. https://doi.org/10.1016/j.bodyim.2010.11.001
  8. Coetzee V, Perrett DI, Stephen ID. Facial Adiposity: A Cue to Health? Perception. 2009. https://doi.org/10.1068/p6423

← All references