How old do I look? Why age guessers disagree, and what ages a face

Table of contents

• Key numbers
• How old do I look — can anything actually tell me?
• How accurate are AI age guessers?
• How accurate are people?
• Why do five tools give five different ages?
• Which age guessers are out there, and what do they claim?
• What actually makes a man's face read older?
• What can you change in a month, and what takes years?
• The bottom line
• Studies referenced

The first site says 38. You are 29. A second tab says 26, then a third says 33, and suddenly one face has acquired twelve years of disagreement. You are not imagining the inconsistency: each tool is reading a different photograph through a different model.

The direct answer is that no age guesser can tell you your exact age from one image. A result is a rough band, not a verdict. If what you really want to know is how your face lands on a stranger, that is a different question — and a more useful one.

TL;DR

• No tool gives you one true age: The best algorithms miss by about three years on clean ID photos, and by more on selfies.
• Expect disagreement: AI and humans both miss by years, especially when the photo is smiling, close, rotated, low-contrast or wearing glasses.
• Read the input: Sun exposure and smoking have the strongest evidence behind an older facial read; sleep loss, hair and a gaunt older face add different cues.
• Use a number carefully: A guess tells you how one image leaned, not what age strangers will always see.
• If the real question is how you come across: Our free photo perception test does not output an age; it estimates how your photos read to a stranger. It is a perception test, not a clinical instrument.

Key numbers

• 100 milliseconds: Trait judgments can form after a very brief face exposure (Willis & Todorov, 2006).
• 3.1 years: The best mean absolute error in NIST's common visa-photo evaluation (Hanaoka et al., 2024).
• 5.15 years: The median inaccuracy for undisguised faces in one study (Thorley et al., 2022).
• 2.4 versus 1.1 years: Smiling made young-adult faces read older by this amount for AI versus humans (Ganel, Sofer & Goodale, 2022).
• 24% less: Skin-surface ageing in the daily-sunscreen group over 4.5 years (Hughes et al., 2013).

How old do I look — can anything actually tell me?

A tool or stranger can place you in a band about ten years wide, not a precise age: ask which way the photo leans and why. A photo has an age; you have a range.

If the question underneath is how you come across, our free perception test estimates how your photos read to a stranger on a 70–155 perception axis. It does not guess your age, and it is a perception test, not a clinical instrument. There is no paywall on your score; opening the result needs a free account.

How accurate are AI age guessers?

The best systems work within several years on clean ID photos, then degrade on ordinary images. NIST's 2024 evaluation of six prototypes found the best visa-photo mean absolute error improved from 4.3 to 3.1 years; border webcams were hardest for four of six algorithms, and four of six erred more with glasses. Error was generally higher in women.

The NIST evaluation did not measure skin tone, so it says nothing either way about skin-tone bias — and none of it was tested on bathroom selfies.

A study of 21 AIs found humans more accurate overall, but AI beat the human sample on middle-aged male faces (4.62 versus 6.44 years); on faces aged 60–80, AI error was 7.05 versus 6.77. The human sample was 30 undergraduates; the 2022 comparison found young smiling faces difficult for AI and wide variation between systems.

In 2015, Microsoft's engineers said how-old.net “isn't particularly accurate with age, but it's often good for a laugh.” Their own write-up says they expected about 50 test users and got over 35,000.

A cleaner benchmark can tell us how a model performs on that benchmark; it cannot turn a messy selfie into a birth certificate.

How accurate are people?

People are noisy age estimators. A passport-photo study found average error “approximately 8 years” on a root-mean-square basis; another found undisguised-face estimates inaccurate by a median of 5.15 years. “You look 35” is an impression, not a measurement.

Younger faces are pushed older, older faces younger, and estimates drift toward the preceding face. So says the passport-photo research.

Expression evidence is mixed: one study found happy faces underestimated, another found smiling faces read older. Either way it moves a guess by a year or two. Reading older is not automatically bad, as our guide to the most attractive age for a man explains.

Photo: Vitaly Gariev on Pexels

Why do five tools give five different ages?

Different training data, and a photo that varies more than you think. A smile, a pair of glasses, flat light or a close lens each shift the estimate, so five uploads are really five different inputs.

• Smile: Ganel found smiling faces read older: about 2.4 years for the AIs against 1.1 for humans, on young-adult faces.
• Glasses: NIST found higher error for four of six algorithms. For humans, the effect is “rather small”; “five years older” comes from a clinic survey, not a study.
• Contrast: Higher facial contrast read younger, lower contrast older (Porcheron et al., 2017).
• Distance: Close-range photos received lower trustworthiness, competence and attractiveness ratings (Bryan, Perona & Adolphs, 2012), but the study did not measure age.

Check the photo before you diagnose the face: see photo lighting and angle and front camera versus back camera.

The studies do not all point in the same direction on smiling, which is a reason to report uncertainty rather than manufacture a single rule.

Which age guessers are out there, and what do they claim?

All of them are free to try. They differ in what they claim and in what they want back. Treat accuracy wording as product copy, not a shared scientific standard.

faceage.ai and NOVOS “FaceAge” share a name with the 2025 Lancet Digital Health FaceAge study, but neither page mentions it. The FaceAge study trained on presumed healthy people aged 60 or older and was tested on cancer patients; it does not validate a consumer selfie tool.

We do not make an age guesser. A chatbot is not one either, for the reasons in can ChatGPT rate your face.

What actually makes a man's face read older?

Sun exposure and smoking have the strongest evidence; sleep loss produces tired cues, hair changes the social read, and very low body weight can age an older face. None gives a universal “years older” conversion; some measure skin or tiredness rather than perceived age.

• Sun: A randomised trial found daily sunscreen reduced skin-surface ageing by 24% over 4.5 years; it did not ask raters how old faces looked. Another estimate said UV exposure “seems to be responsible for 80%” of visible facial-ageing signs in 298 Caucasian women. Read sunscreen every day as prevention, not a younger-number promise.
• Smoking: In 79 twin pairs, smokers scored worse on eyelid skin, bags, folds, wrinkles and jowls; no “years older” figure exists in the abstract. Among Danish twins aged 70 and over, older appearance went with smoking, sun exposure and low BMI; about 40% of perceived-age variation was non-genetic. See under-eye bags and first impression.
• Sleep: Sleep-deprived faces were rated less healthy, more tired and slightly less attractive, with hanging eyelids, dark circles and paler skin. Nobody rated their age. See sleep and appearance statistics.
• Hair and weight: Removing hair digitally added “almost 4 years” in a study of four men, but a second real-shaving study found no significant age difference. In men over 70, low BMI went with looking older. Compare does grey hair make you look older with should I shave my head.

An age ten years too high from a free web tool is one noisy reading of one photo. Before it sends you shopping for procedures, take three more in daylight.

Perceived age does have health research behind it: among 1,826 twins aged 70 and over it predicted survival after adjusting for real age. That is a finding about older populations, not evidence that an online guess predicts your health or your dating life. If the tools keep guessing you too young instead, see why do I look younger than my age.

What can you change in a month, and what takes years?

In a month you can change what the photo shows: sleep, haircut, facial hair, glasses, expression and light. Over years you protect skin and do not smoke. These are habits, not guarantees.

In a month

• Stabilise: Get regular sleep; tired cues show.
• Groom: Test a haircut and facial-hair shape that frames the face you have now.
• Photograph: Use even front light, a sensible distance and a neutral expression.

Over years

• Protect: Daily sunscreen has trial evidence for less skin-surface ageing; see how to age well as a man.
• Avoid: Smoking and heavy sun exposure are the clearest avoidable risks.
• Stay steady: When older, do not chase extreme leanness if it makes the face gaunt.

The bottom line

An AI age guesser can entertain and sometimes inform, but its answer is the midpoint of a broad band, and people are not much cleaner. Different outputs mean different models, photos and biases — not a secret age your face is hiding.

A photo has an age. You have a range. Look for the direction your photos lean and the cues you can control. Our free photo perception test does not output an age; it estimates how your photos read to a stranger on a 70–155 perception axis. It is a perception test, not a clinical instrument; it has not been validated against real-world outcomes.

Studies referenced

• Axelsson, J., Sundelin, T., Ingre, M., Van Someren, E. J. W., Olsson, A., & Lekander, M. (2010). Beauty sleep: experimental study on the perceived health and attractiveness of sleep deprived people. BMJ, 341, c6614. doi:10.1136/bmj.c6614
• Bontempi, D., Zalay, O., Bitterman, D. S., et al. (2025). FaceAge, a deep learning system to estimate biological age from face photographs to improve prognostication: a model development and validation study. The Lancet Digital Health, 7(6), 100870. doi:10.1016/j.landig.2025.03.002
• Brunet, N. M., & Sharp, J. (2020). Do glasses modulate age perception? i-Perception, 11(4), 2041669520953457. doi:10.1177/2041669520953457
• Bryan, R., Perona, P., & Adolphs, R. (2012). Perspective distortion from interpersonal distance is an implicit visual cue for social judgments of faces. PLoS ONE, 7(9), e45301. doi:10.1371/journal.pone.0045301
• Christensen, K., Thinggaard, M., McGue, M., Rexbye, H., Hjelmborg, J. v. B., Aviv, A., Gunn, D., van der Ouderaa, F., & Vaupel, J. W. (2009). Perceived age as clinically useful biomarker of ageing: cohort study. BMJ, 339, b5262. doi:10.1136/bmj.b5262
• Clifford, C. W. G., Watson, T. L., & White, D. (2018). Two sources of bias explain errors in facial age estimation. Royal Society Open Science, 5(10), 180841. doi:10.1098/rsos.180841
• Flament, F., Bazin, R., Laquieze, S., Rubert, V., Simonpietri, E., & Piot, B. (2013). Effect of the sun on visible clinical signs of aging in Caucasian skin. Clinical, Cosmetic and Investigational Dermatology, 6, 221–232. doi:10.2147/CCID.S44686
• Ganel, T. (2015). Smiling makes you look older. Psychonomic Bulletin & Review, 22(6), 1671–1677. doi:10.3758/s13423-015-0822-7
• Ganel, T., Sofer, C., & Goodale, M. A. (2022). Biases in human perception of facial age are present and more exaggerated in current AI technology. Scientific Reports, 12, 22519. doi:10.1038/s41598-022-27009-w
• Hanaoka, K., Ngan, M., Yang, J., Quinn, G. W., Hom, A., & Grother, P. (2024). Face Analysis Technology Evaluation: Age Estimation and Verification (NIST IR 8525). National Institute of Standards and Technology. doi:10.6028/NIST.IR.8525
• Hughes, M. C. B., Williams, G. M., Baker, P., & Green, A. C. (2013). Sunscreen and prevention of skin aging: a randomized trial. Annals of Internal Medicine, 158(11), 781–790. doi:10.7326/0003-4819-158-11-201306040-00002
• Mannes, A. E. (2013). Shorn scalps and perceptions of male dominance. Social Psychological and Personality Science, 4(2), 198–205. doi:10.1177/1948550612449490
• Okada, H. C., Alleyne, B., Varghai, K., Kinder, K., & Guyuron, B. (2013). Facial changes caused by smoking: a comparison between smoking and nonsmoking identical twins. Plastic and Reconstructive Surgery, 132(5), 1085–1092. doi:10.1097/PRS.0b013e3182a4c20a
• Porcheron, A., Mauger, E., Soppelsa, F., Liu, Y., Ge, L., Pascalis, O., Russell, R., & Morizot, F. (2017). Facial contrast is a cross-cultural cue for perceiving age. Frontiers in Psychology, 8, 1208. doi:10.3389/fpsyg.2017.01208
• Rexbye, H., Petersen, I., Johansens, M., Klitkou, L., Jeune, B., & Christensen, K. (2006). Influence of environmental factors on facial ageing. Age and Ageing, 35(2), 110–115. doi:10.1093/ageing/afj031
• Sundelin, T., Lekander, M., Kecklund, G., Van Someren, E. J. W., Olsson, A., & Axelsson, J. (2013). Cues of fatigue: effects of sleep deprivation on facial appearance. Sleep, 36(9), 1355–1360. doi:10.5665/sleep.2964
• Thorley, C., Acton, B., Armstrong, J., Ford, S., & Gundry, M. (2022). Are estimates of faces' ages less accurate when they wear sunglasses or face masks and do these disguises make it harder to later recognise the faces when undisguised? Cognitive Research: Principles and Implications, 7(1), 17. doi:10.1186/s41235-022-00370-0
• Voelkle, M. C., Ebner, N. C., Lindenberger, U., & Riediger, M. (2012). Let me guess how old you are: effects of age, gender, and facial expression on perceptions of age. Psychology and Aging, 27(2), 265–277. doi:10.1037/a0025065
• Willis, J., & Todorov, A. (2006). First impressions: making up your mind after a 100-ms exposure to a face. Psychological Science, 17(7), 592–598. doi:10.1111/j.1467-9280.2006.01750.x
• Tool claims and prices were read from each tool's own public page on 19 September 2026 and may change. We did not upload photos to any of them.