AI attractiveness test that tells you the real number

What a single photo can and cannot tell you

Every AI attractiveness test works from the same input: one 2D photograph. That input decides what the test can honestly claim. It can tell you how a face reads, which is what attractiveness is: a perception, formed by other people in a fraction of a second, from exactly this kind of flat image. It cannot tell you the geometry underneath.

So BecomeTen does not report measurements. There is no angle in degrees, no ratio to two decimals, no millimetre. Each structural trait comes back as a directional band: canthal tilt positive, neutral or negative; midface ratio shorter or longer; gonial angle sharper or softer. A tool that reads a precise number off a selfie is inventing the decimals, and the people who know the vocabulary can tell.

What the photo does support is a structured opinion against a fixed rubric across 30 or more traits in five areas: harmony, eye area, jaw and chin, skin, hair. That opinion is the score.

What does the research say actually drives attractiveness ratings?

The forum version of attractiveness is a long list of measurements. The research version is shorter, less precise and better supported, and it is the version the rubric is built on.

• Symmetry and averageness. A 1994 study found that both facial symmetry and closeness to the population average correlate with rated attractiveness, and that each does so independently of the other.
• Skin homogeneity. A 2007 study of female facial skin found that evenness of colour, with face shape held constant, changes how old, how healthy and how attractive a face is judged.
• Facial adiposity. A 2009 study found that the amount of fat visible in a face is a cue observers use to judge health, and that it tracks rated attractiveness across the normal range.

Weighting follows how faces are actually read: harmony carries the most, then the eye area, then jaw and chin, then skin and hair. Skin is the most improvable area, not the most defining one.

Why do lighting and angle move the score?

Because the test rates the photo, and the photo is not the face. A camera flattens three dimensions into two, and the way it does so depends on where the light is, where the lens is and how far away it sits. Change any of the three and the same face produces a different image with different trait bands.

• Angle. A chin tilted down sharpens the jaw and shifts the whole lower third; a chin tilted up widens the midface and shortens the nose. This is the single most common false positive.
• Light. Overhead light hollows the eye sockets and darkens the under-eye; front light fills them. Harsh light exaggerates texture; soft light hides it.
• Distance. A close, wide phone lens enlarges whatever is nearest, usually the nose, and narrows the sides of the face.
• Expression and filters. A smile lifts the cheeks and narrows the eyes; a beauty filter removes the skin signal the test reads.

Every photo is graded for light, angle and sharpness before it is scored; a poor capture is flagged, not quietly scored, and comparisons against it are withheld. The full rules are in photo quality: light, angle and lens.

What does potential mean, and where does it stop?

Potential is the score reachable without surgery. It is not a promise and not a prediction; it is a ceiling, worked out per area and clamped in code so the model cannot exceed it. The caps follow what genuinely moves: harmony up to 6 points, the eye area up to 10, jaw and chin up to 14, skin up to 25, hair up to 20. Bone geometry does not change, which is why harmony, the area most tied to bone, gets the least.

Every weak point in the report is tagged workable or structural. For a structural one the plan will never claim a fix; it may only work with how the trait reads through composition, posture, hair framing and grooming. The looksmaxxing guide walks through each area and what has evidence behind it.

What is a good attractiveness score?

Read the score against the calibration, not against a ten-point instinct. An average, healthy, groomed adult lands between 50 and 60, the Normie tier, and on this scale that is a good score in the plain sense: nothing is badly wrong. Below 50, one or two areas are usually pulling the rest down, and in most photos those areas are workable ones.

Beside the score you get the PSL equivalent, so the number also reads in the forum's vocabulary: 50 is about PSL 4, 62 about 5, 75 about 6. The full tier ladder is on the AI face rating page, and the mapping is explained on PSL score.

What this test refuses to do

• No leaderboard. You are not ranked against other users, and the percentile shown is a position on the scale, never a position among people.
• No ethnicity or gender adjustment. The rubric is applied to the photo in front of it, the same way for every face. There is no separate scale by group.
• No surgery recommendations, at any tier. The plan works only with what is reachable without a scalpel.
• Nothing that damages tissue. Bone smashing and similar forum methods are forbidden in the prompt that writes your plan.
• No demeaning labels. The tiers are neutral bands on our own scale, not the community's ladder of insults.
• No diagnosis. The rating is an estimate for educational use, not a medical opinion.

Each of those has a cost. A leaderboard would drive sharing; a promise about bone would sell more reports. The methodology page explains why each is refused anyway.

What is free and what the report costs

The score, tier, PSL equivalent and a share card are free, one scan per device per day, no sign-up, no card, photo deleted right after scoring. The full report is a one-time €6.99 and adds the 30+ trait bands, the weak points with their workable or structural tag, the potential ceiling and gap, a plan with an evidence level on every step, and an AI potential photo.

A re-scan is €4.99. BecomeTen PRO is €9.99 a month or €59.99 a year: 20 scans per rolling 30 days, at most 3 a day, each with its potential photo, plus progress tracking. Payment comes after the score, and the refund terms are public.