Do I look my age

Three rows, and how many years

The report shows one integer, the age you look in this photo, and a likely range from a low year to a high year. The integer always sits inside that range. Your birthday is not an input. You bring it yourself and compare it with the range, not with the integer alone.

The ends count as inside. A birthday equal to the low year or the high year is still your age for this photo. Years older, or years younger, means the gap between your birthday and the nearest end, not the gap to the integer.

Suppose the report says you look 34, with a likely range of 31 to 37. That pair of numbers is an illustration of the reading rule, not a result from a real photo.

• Your age inside the range. A real age of 31, 34, or 37 is inside. You look your age in this photo, even when the integer is 34 and the birthday is 37.
• Your age above the range. A real age of 40 is three years above the high end. The face reads about three years younger than you are in this photo.
• Your age below the range. A real age of 28 is three years under the low end. The face reads about three years older than you are in this photo.

How old do I look compared to my age

Use the range as the answer and the integer as the single best guess inside it. A birthday one year off the integer, still inside the range, is your age for this photo. Treating that one-year gap as a verdict throws away the line the report is built around.

Confidence is low, medium, or high. High is a sharp, front-facing photo in even light. Low means blur, light, angle, a filter, or makeup limited what could be seen, and the instructions ask for a wider range in that case. The page still shows an estimate. A low-confidence range that contains your birthday is a softer yes than a high-confidence one. If the badge says photo quality limited the estimate, take another photo before you decide anything.

The instructions ask for a range that is usually 4 to 8 years wide. The page then keeps the span honest. A span wider than 20 years is replaced with 10 years on either side of the integer. A span narrower than 2 years is replaced with 1 year on either side. What you see is therefore 2 to 20 years wide. A range of one year would claim a precision the photo does not have.

A match is a band of years

Human raters miss by several years even on neat passport photos, so looking your age was never a single year. Clifford, Watson, and White, Royal Society Open Science, 2018 tested 3,948 standardized passport images with 84 perceivers across three experiments. The abstract puts average error at about 8 years. In experiment 1, root-mean-square error on the unmasked photos was 8.36 years (95% CI 7.70 to 9.03). Masking the face raised it to 9.64 years, then 10.79 years. Those are sizes of the miss, not a constant shift.

Two biases did most of the explaining, together about 95% of the variance in their model. The larger one is serial: each guess was pulled toward the age of the face shown just before. The smaller one is a slope below 1. Across stimulus ages 7 to 70, overall bias was -0.46 years and not distinguishable from zero. The slope corresponds to a bit over 1.5 years: the youngest faces judged slightly older, the oldest slightly younger. An 8-year miss is mostly scatter and the previous face, not a rule that every young adult is seen as a decade older.

Voelkle, Ebner, Lindenberger, and Riediger, Psychology and Aging, 2012 asked 154 adults to estimate 171 faces on 2,052 photographs from the FACES database. Mean absolute error was 5.91 years for young perceivers (95% CI 5.59 to 6.22), 6.30 for middle-aged perceivers, and 6.83 for older perceivers. Older faces were harder (7.40 years) than young faces (5.59). Neutral expressions were the most accurate. Happy expressions were the ones most often judged younger than the true age. People were more accurate on faces near their own age, and there was no reliable own-gender advantage.

A stranger, a friend your own age, and this report can disagree by several years and all sit inside those published errors. A gap smaller than that band is ordinary. It is not a health finding.

What people say about themselves is a third number, and it is not a stranger's rating. A 2024 paper in Psychology and Aging used the December 2019 National Poll on Healthy Aging: 2,006 U.S. adults ages 50 to 80. Asked how they look next to people their age, 59% said younger, 35% said the same, and 6% said older. Most adults in that sample do not claim to look their age. That is a self-description. It does not mean a rater, or this report, would place 59% of those faces under the birthday.

A two-photo test

One photo is one setup. Take a second photo and read the overlap. The shooting rules live on the face age photo checklist: one face, the original file, front light, eye-level distance, a relaxed expression, no sunglasses and no beauty filter. This page does not repeat that list.

Upload the two photos as two separate estimates. A file with two clear faces is refused, so a side-by-side with a friend is not a comparison. Crop to one person first.

Your birthday inside the overlap of the two ranges means both photos say you look your age. Inside only one range means the photos disagree. Light and expression are the first place to look. In the Voelkle study a smile was the expression most often rated younger, so a grin in one frame and a neutral mouth in the other can move the range without any change in your face across the year.

The serial effect in Clifford's passport study is a human effect: the previous face pulls the next human guess. Each estimate here is one photo, and the request is not shown the face you looked at before it. A quiz that flashes many ages in a row does have that pull. That is one reason a friend quiz and this report can disagree on the same picture.

If the two ranges do not overlap each other at all, do not average the integers and call that your age. The photos are not the same input. Match the checklist and run the pair again.

A quiz of friends

A quiz that asks people for a single year answers a social question. Friends round toward a kind number. Write the guesses down and look at the spread. If the guesses run from 29 to 37, a tidy average of 33 hides an 8-year band, which is about the size of the passport-photo error above.

Someone your own age is a bit closer, on the Voelkle results, and still misses by about 6 years. A stack of celebrity faces makes the next human guess worse, because of the serial pull. This photo test does not collect other people's guesses, and it does not ask you to type a birthday.

Pass the same photo to a friend after you have both looked at the range, not before. You are then comparing a number with a band, which is the fair version of the quiz.

What the report gives you

You get the integer, the range, a confidence level, 3 to 5 cues marked older, younger, or neutral, and 2 notes. It is free, with no account, and 10 estimates a day. JPG, PNG, and WebP are accepted, up to 25 MB. The browser resizes the photo before upload. The photo is analyzed in memory and never stored. The privacy policy is the full path.

No face, a face too small to judge, or features that cannot be made out at all comes back as a message instead of a number. A slightly soft or dim photo still gets an estimate, with confidence low. Several faces are refused as a set. There is no per-face age on a group photo.

The gap is not a medical result, not a biological age, and not a check you can show at a door. Causes that have published sizes, such as sun, smoking, facial fat, and feature contrast, are on why a face reads younger than the birthday. Beard, glasses, and makeup, and which of those shifts a study actually measured, are on how to look older. Cue by cue on a single photo, start at How Old Do I Look AI. What the same photo can and cannot say about skin is the skin age test.

Questions about the comparison