Finally — a personality test that measures you instead of asking.

120 facial landmarks. 120,000+ profile training data. Published variance on every trait. (Beyond MBTI. Beyond ChatGPT.)

You just read the article — here's the scan that measures instead of asks.

How It Works

Upload a photo of yourself. Front-facing, neutral expression. 5 seconds.

120-landmark measurement. Orbital width, philtrum depth, jaw angle, 117 more

all measured against 120k+ profile training data.

Read the scorecard. Per-trait scores with published variance. Plus the Five-Element overlay and a Compatibility Matrix you can save.

What You Get

Trait scoring you can argue with (not negotiate with)

Five-Element archetype (Wood/Fire/Earth/Metal/Water) cross-referenced with traits

A "hidden blind spots" section that names what self-report misses

A 100-Year Life Map with predicted developmental windows

Per-trait variance published on every score (most tests skip this entirely)

Common Questions

"Pseudoscience / Barnum effect."

We publish per-trait variance. The methodology page lists every landmark and the reference baseline. That's a higher bar than most personality assessments meet.

"ChatGPT could already do this."

ChatGPT is a general LLM. SomaScan is trained on 120,000+ profiles with stated reference baselines for facial-landmark-to-trait correlations. Different mechanism, different output.

"What's measurable about a face?"

Orbital width, philtrum depth, jaw angle — these are observable, replicable measurements. The methodology page lists all 120 landmarks.

Is this Barnum-effect or actual measurement?

Every trait carries a published variance score. Run it on someone you've known 20 years to test.

Why is it different from MBTI?

MBTI is self-report. SomaScan is measurement of facial landmarks. Different data input, different reliability.

What about people who don't look like their personality?

That's exactly the variance the report shows. Mismatch is signal, not noise.

Is this evidence-based?

Methodology page lists the 120 landmark-to-trait correlations and the variance per trait. Higher transparency than the category norm.

What if I scan myself and the result feels wrong?

Read it with the variance score. If variance is high, the trait read is uncertain. Most readings have at least one high-variance trait — that's where the blind spot usually lives.