DEXA vs. Calipers vs. BIA vs. Photo: How Accurate Is Each Body Fat Method?

The short answer: DEXA is the most accurate of the four common body-fat methods, but none of them — DEXA included — measures fat directly. For tracking at home, the most useful method isn’t the one with the best lab accuracy; it’s the one you can repeat under identical conditions, because a consistent method shows a real trend even when its absolute number is off. If you only have a tape measure or calipers, the free body fat calculator runs the standard equations.

Every body fat method has a sales pitch. Calipers are “what bodybuilders have always used.” Smart scales promise a number every morning. DEXA gets called the gold standard so often that people forget it’s an estimate too, just a very good one. None of them measure fat directly the way a scale measures weight. Every single one is inferring fat percentage from some other signal, X-ray attenuation, a pinch of skin, electrical resistance, or now, visual pattern recognition from a photo. That’s worth knowing before you trust any of them to the decimal.

DEXA: the reference method, with real limits of its own

A DEXA (dual-energy X-ray absorptiometry) scan estimates body composition by passing two X-ray energies through the body and measuring how differently bone, fat, and lean tissue absorb them. It’s the method most other methods get validated against, and for good reason: it’s not affected by hydration status the way bioelectrical methods are, and it produces a full breakdown, not just one number.

It is not flawless. Different DEXA machines and software versions can produce meaningfully different results for the same person, hydration still shifts soft-tissue readings somewhat even though less than BIA, and a scan only costs $50 to $150 and requires a clinic visit, which makes weekly or even monthly scanning impractical for almost everyone. Repeat scans on the same machine are consistent, but a DEXA number is still an estimate built on assumptions about tissue — very good, not infinite.

Skinfold calipers: cheap, but entirely technician-dependent

Calipers measure the thickness of a pinch of skin and underlying fat at specific sites, then run that measurement through a population-based equation to estimate total body fat. The core limitation is structural, not just about skill: calipers only capture subcutaneous fat, the layer directly under the skin, and have no way to account for visceral fat, the fat around internal organs. Someone carrying more visceral fat than the equation assumes will read as leaner than they actually are.

On top of that, measurement error between different people doing the pinching, or even the same person measuring inconsistently, routinely runs several percentage points. Calipers can be a reasonable trend tool if the same trained person measures you the same way every time. They’re a poor tool for a one-off number.

Bioelectrical impedance (smart scales): convenient, systematically biased

BIA sends a tiny electrical current through the body and estimates fat percentage from how much resistance it meets, since fat and lean tissue conduct electricity differently. It’s the technology behind most consumer smart scales, which is why it’s the most common home method by far.

The tradeoff is that BIA is highly sensitive to hydration, meal timing, recent exercise, and even room temperature, all things that change your body’s conductivity without changing your actual fat mass. Research comparing BIA against DEXA consistently finds it underestimates or overestimates depending on the population and the specific device, not in one predictable direction across everyone. A smart scale used at the exact same time of day, in the exact same state, can still be a decent trend tool. Used at random times, its number-to-number variation can rival the actual change you’re trying to detect.

Two studies show the pattern clearly. In a 2021 comparison against DEXA, three consumer smart scales measured body weight to within a median of 0 to 0.3 kg, but underestimated fat mass by a median of 2.2 to 4.4 kg. A 2023 study of fifteen BIA devices found every one highly repeatable (precision error under 0.5%), yet single-reading error ranged from −3.5 to +11.7 percentage points depending on the device. That’s the definition of consistent but biased — useful for a trend on one device, unreliable as an absolute number.

Photo-based estimates: no equipment, but capture-dependent

Photo-based estimation is the newest of the four. A model trained on large sets of photos paired with reference measurements learns to read visual markers, muscle definition, how fat sits over specific areas, overall shape, and map them to an estimated body fat percentage. How this actually works, and how Corpus does it specifically, is worth reading in full if you’re evaluating it.

The upside is real: no equipment, no clinic visit, and it’s the only one of the four that can double as a visual record you can look back at directly. The downside is that quality depends entirely on capture consistency, framing, distance, lighting, and pose all shift what the model sees, the same weakness that affects a trained human eye looking at two photos. A single well-lit, correctly-framed photo carries meaningfully more signal than a dim, distant, badly-posed one. A 2026 review of eleven research-backed body-composition apps found good group-level accuracy for body fat percentage, but reduced accuracy for individuals and at higher body fat, and limited evidence so far on tracking change over time.

So which one should you actually use?

Realistically, most people should think about this in two tiers. If you want the closest thing to a lab-grade number once or twice a year, DEXA is worth the cost and the drive. For the other 363 days, when a lab visit isn’t realistic, the honest answer is: whichever method you’ll actually use consistently, under matched conditions, read as a trend rather than a verdict. That’s the entire premise behind why your weight changes before your body does, and it applies just as much to body fat percentage as it does to scale weight.

The takeaway

There is no method here that hands you a precise number with no strings attached. DEXA is closest, and least practical for regular use. Calipers and smart scales are practical, and each carry their own systematic bias. Photo-based estimates trade equipment for a dependency on consistent capture. The real question isn’t “which one is right,” it’s “which one will I actually repeat the same way, often enough, to see a real trend.” See Corpus’s full methodology for exactly how the photo-based estimate is built and where its limits are.