How accurate is a photo body scan, honestly

Every number is an estimate

No tool you can buy measures body fat. Every one of them estimates it. DEXA infers fat from how two X-ray energies pass through tissue. Calipers feed a fold of skin into equations from the 1970s. A smart scale passes a current through you and measures the resistance. A photo scan reads your outline. The lab reference — the 4-compartment model — is the only near-truth, and it lives in research facilities.

Judged against that reference, even DEXA lands 1–2 body-fat percentage points off for an individual — sometimes more. So the honest question is never "is this accurate". It is "how wrong is it, typically — and does the wrongness repeat".

The error bars, side by side

Typical individual error against a lab reference: DEXA, ±1–2 points. Skinfold calipers in trained hands, ±3–5 — worse in untrained hands, and yours are untrained. Consumer BIA scales, ±3.5–8 depending on the model and your hydration. Photo scans, roughly ±2.5–5 in validation work. Studies consistently find BIA drifts most, because water dominates impedance and water moves hourly.

Read the overlap honestly. A photo scan sits in the caliper-and-scale band, not the DEXA band. If you need one clinically defensible number, book a DEXA. If you need a number every month, at home, under conditions you control, the photo scan's band is workable — for one specific job.

What a photo scan actually does

A photo scan maps your silhouette and landmarks, then a model trained on paired scans estimates composition and circumferences. Its error is partly systematic: it may read you 3 points high every single time. A repeating bias cancels out when you compare this month to last month. That is the property that makes trends usable. What it cannot see: visceral fat, bone density, anything under the skin.

The rest of the error is yours to control. Same lighting, same distance, same pose, same tight clothing, same time of morning. Change any of those and the scan measures your conditions, not your body.

Use it for the trend

One scan is a snapshot with a ±3-point shadow. Twelve monthly scans under fixed conditions are a line, and a line can outrun its own error. A waist estimate that falls 2 cm and keeps falling across three scans is real change; a 0.5-point body-fat drop in a single reading is noise — from every tool on this page. Read direction across scans. Let the decimals go. One caveat: illness, a salty week, or the water-retaining phase of a menstrual cycle can shift any of these readings for days — compare like with like.