Is an InBody scan accurate?

It correlates well with DEXA at the population level (r = 0.89 to 0.92) but underestimates body fat percentage by 2.8 to 4.2 percentage points on average, with wide individual-level limits of agreement — in one 1,000-person study, an individual man's result could plausibly be off by up to 10 points in either direction.

Should I trust the exact body fat percentage InBody gives me?

Treat it as an estimate, not a diagnosis. The number is biased in a fairly consistent direction (it tends to underestimate fat and overestimate lean mass), so it is more useful for comparing your own readings over time than for matching an absolute number to a health chart.

Can InBody track my progress accurately even if the number is biased?

Yes, in the one study that tested this directly. Over a 4-week diet, the change InBody measured was not statistically different from the change DEXA measured, even though InBody's starting and ending numbers were both offset from DEXA's. Bias in the absolute reading did not prevent it from tracking the trend.

Is InBody Accurate? What the Validation Studies Show

InBody scanners sit in gyms, clinics, and physio offices, and the number they hand you after 60 seconds standing on a plate feels precise: one decimal place, broken down by segment, printed on a receipt. Precise is not the same as accurate. Two independent validation studies against DEXA — the imaging-based reference standard — show a consistent, well-quantified gap, and a more useful finding underneath it.

The short answer

InBody correlates reasonably well with DEXA at the population level: r = 0.89 in men, r = 0.92 in women, across 1,000 people.
It underestimates body fat percentage by 2.8 to 4.2 percentage points on average, and overestimates fat-free mass to match.
The individual-level spread is wide — for men, a single reading could plausibly sit anywhere from 10 points too low to 1.5 points too high compared to DEXA.
Despite that bias, InBody tracked a real 4-week change in body fat percentage just as well as DEXA did. The offset doesn't cancel out its usefulness for watching a trend.

What a large validation study actually found

The most direct evidence comes from a 2025 study in European Journal of Clinical Nutrition, which compared InBody 770 (a multi-frequency BIA device — the same underlying technology most InBody models use) against DEXA in 1,000 healthy adults: 667 men and 333 women.

For percent body fat, the correlation was r = 0.89 in men and r = 0.92 in women. That sounds strong, and at a glance it is — but correlation measures whether two methods move together, not whether they agree in absolute terms. The study's own concordance figures, which do account for agreement, were noticeably lower: 0.73 in men and 0.84 in women (study, via PMC).

The bias was consistent in direction: InBody underestimated body fat percentage by 4.2 ± 3.0% in men and 2.8 ± 2.6% in women, while overestimating fat-free mass to compensate. The authors were direct about the individual-level implication — the limits of agreement for men spanned −10.01 to +1.57 percentage points. That is the range a single reading could plausibly fall within relative to a DEXA scan on the same day. A 22% reading could, within that range, correspond to a true value anywhere from about 12% to 24%.

The paper's own conclusion is worth stating plainly rather than summarizing: multi-frequency BIA "can provide a reliable and accurate method for assessing common regional and whole-body" composition in real-world conditions — while explicitly noting it "underestimate[s] FM and over-estimate[s] FFM" relative to DEXA. Both statements are true at once. Reliable does not mean unbiased.

The more useful finding: tracking change over time

A separate, smaller study asked a more practical question: if InBody's absolute number is biased, does it still detect change correctly?

Researchers measured 41 exercise-trained adults (29 women, 12 men) with both DEXA and InBody 770 before and after a 4-week calorie-deficit diet. At baseline, InBody read about 3.9 percentage points lower than DEXA for body fat — consistent with the larger study's bias. But when the researchers compared the change each method measured over the 4 weeks, there was no statistically significant difference: DEXA measured a change of −1.3 ± 0.9 percentage points; InBody measured −1.4 ± 1.8 (p = 0.18) (study, via PMC).

The paper's conclusion: "when tracking body composition over a period of four weeks, the MF-BIA may be a viable alternative to the DXA in exercise-trained men and women." The authors also flagged that InBody's individual changes were noisier — a real caveat — but the mean trend held up.

This is the same pattern that shows up across every body-composition method covered on this site. AI photo-based estimates have their own bias and failure conditions. The scale and the tape measure answer different questions and neither is perfectly precise on a single day. A biased instrument that is consistently biased in the same direction can still be a reasonably good change detector, because the bias mostly cancels out when you subtract one reading from another — it's the absolute number that stays wrong.

What this actually means if you use InBody

Don't compare your InBody number directly to a DEXA-based health chart. The 35%/40% waist-circumference-style thresholds people quote for body fat percentage are typically built on DEXA or similarly precise data. Plugging an InBody reading straight into one of those charts inherits whatever bias your specific reading happens to carry.

Do compare an InBody reading to your own past InBody readings, ideally from the same machine, at a similar hydration state and time of day — bioelectrical impedance is sensitive to water content, the same way a bathroom BIA scale is. Consistency of conditions matters more than which device you use.

Treat a single scan as one data point, not a verdict. The individual-level spread in the 1,000-person study — up to 10 percentage points for some men — is too wide to hang a decision on one reading. A pattern across several scans, weeks apart, carries far more signal than any single number, however precisely it's printed.

Bottom line

InBody is a legitimate, validated tool at the population level, and the research doesn't support dismissing it as inaccurate. It's also not DEXA, and the studies are specific about the size and direction of that gap: it tends to underestimate fat and overestimate lean mass by a few percentage points, with individual readings varying more than the average bias would suggest. What it's shown to do well is track a real change over weeks — which, for almost everyone using it, is the actual question they're asking.