Body Scanner Comparison: AI Scan vs DEXA vs InBody

Step on a scale and you learn one number. It does not tell you whether the change since last month was fat, muscle, or the extra litre of water you are carrying today. Every method built to answer that better question makes a trade, cost against precision, or precision against convenience, and not much of the marketing around them says so plainly.

This is the plain version. Five ways to track body composition and shape: what each one actually measures, where it is strong, where it quietly falls apart, and how to pick one you will still be using in a year.

Why track body composition or shape at all

The scale is an outcome metric, and a noisy one. Body weight moves with sodium, glycogen, hydration and bowel timing well before it moves with body fat, so a single weigh-in tells you almost nothing about whether training is working. We have made this case before: the number that matters is the trend, not the snapshot, and that is exactly as true for your body as it is for your training log.

Body composition and shape tracking exist to answer a better question than the scale can: is training actually changing what you are made of, and where. That is worth knowing for reasons beyond appearance. Strength athletes want to know if a bulk is adding mostly muscle or mostly stomach. Endurance athletes chasing a power-to-weight target want to know if a cut is costing them the muscle that produces the power. Almost everyone benefits from a signal that survives a bad night’s sleep or a salty dinner, because that signal is what actually tells you whether to keep doing what you are doing.

The method you choose matters less than most marketing suggests. What matters is whether you will repeat it under similar conditions often enough to see a trend. Keep that question in mind through the rest of this, because it is the one that actually decides which method is right for you.

DEXA: the clinical gold standard

A DEXA scan (dual-energy X-ray absorptiometry) is the reference standard clinical researchers use when they test other methods. Pelaris’s AI scan has not been tested against DEXA, so it is best read as a trend, not a measurement. It passes two low-dose X-ray beams through your body and reads the difference in absorption to split you into three compartments: fat mass, lean mass, and bone mineral content, region by region, arm by arm, leg by leg.

That regional breakdown is what nothing else on this list can match. A DEXA scan will tell you your left leg carries more lean mass than your right, which matters if you are rehabbing an old injury or chasing symmetry. It is also the only method here that gives you a genuine bone density reading, relevant for older athletes or anyone with a family history of osteoporosis.

The trade is cost and access. A scan is a private, out-of-pocket cost at most Australian clinics (check your local clinic’s current price), requires a booking, and involves a small dose of radiation, low enough that it is not a practical concern at normal frequencies, but enough that clinics will not scan you weekly. Most people who use DEXA well treat it as a quarterly or twice-yearly checkpoint, not a training tool. It answers “where am I now, precisely” far better than it answers “what changed since Tuesday.”

Bioimpedance: the scale you already own

Bioimpedance analysis (BIA) is what is running inside a smart bathroom scale and inside the stand-on machine with handlebars at most gyms. It sends a low, imperceptible electrical current through your body and measures the resistance. Muscle and water conduct it easily; fat resists it more. From that single resistance reading, the device estimates body fat percentage, muscle mass and water weight.

The appeal is obvious: cheap, instant, and you probably already own one. The catch is that the resistance reading is extremely sensitive to hydration, and hydration moves constantly. Drink a large glass of water, train, eat a salty meal, or measure first thing in the morning versus after work, and the same body can read several percentage points differently on the same device within the same day. Manufacturers know this, which is why every BIA scale instruction sheet buries a line about measuring at a consistent time under consistent conditions, advice most people never read and fewer follow.

None of that makes BIA useless. It makes it a trend tool with an asterisk. Weigh in at the same time of day, in a similar hydration state, on the same device, and the week-to-week noise mostly cancels out, leaving a usable direction of travel. Compare a single reading against a friend’s different scale, or against your own reading from a different time of day, and you are comparing noise to noise.

3D booth scanners: detailed, but tied to a location

The booth scanner some gyms and specialist studios install has you step onto a platform and rotate while a ring of cameras or infrared sensors builds a 3D mesh of your body. From that mesh it extracts circumference measurements at dozens of points, waist, chest, each thigh, each arm, tracks them over repeat visits, and renders a 3D model you can rotate on screen.

This is the most detailed circumference tracking available outside a tailor’s tape measure, and it is genuinely good at showing shape change over a cut or a bulk in a way a single body fat number cannot. The honest limitation is that it lives wherever the machine lives. You need to go back to that specific gym, studio or clinic, on their schedule, to get a comparable follow-up scan on the same hardware. Travel, switch gyms, or the studio closes, and your history is stranded. It is an excellent tool if you belong to a facility that has one and you will keep going back. It is not a tool you carry with you.

Tape and calipers: free, but only as good as your technique

A tape measure and a set of skinfold calipers cost about as little as tracking gets, and in the hands of someone who has practised the technique, calipers can estimate body fat percentage within a few percentage points of a DEXA scan. That is a real, defensible method, not a myth.

The catch is the phrase “in the hands of someone who has practised.” Caliper technique is a skill: the pinch site, the pinch depth, the angle, and the exact anatomical landmark all move the number, and self-measurement is harder still because you cannot see half the sites you need to pinch. Two measurements taken five minutes apart by two different people, or even the same person on two different days, can disagree by more than the actual change you are trying to detect. Tape measurements are more forgiving, waist and hip circumference are simple to learn, but still drift with exactly where you hook the tape and how snug you pull it.

Free and always available count for a lot, and you can run your waist and height through our Tape Body Fat Calculator right now. If you are disciplined about landmarking the same sites the same way every time, ideally with the same second person doing the pinching, tape and calipers remain a legitimate long-run method. If that discipline is not realistic for you, the technique noise will likely swamp the signal you are trying to read.

AI photo analysis: frictionless, and honest about its limits

AI photo analysis, the method Pelaris uses for body scanning, takes a different trade entirely. You take a front photo, and a side photo if you add one, on your phone or laptop, from the same angles each time, and a model estimates your proportions (shoulders, waist, hips) from what it sees, then classifies your build into a shape archetype and tracks how those proportions move between check-ins. Pelaris maps you to one of six body types, each with its own training guide, from the streamlined Linear build to the Oval build, a fuller midsection with real strength underneath.

The trade it makes is precision for frictionlessness. There is no appointment, no travel, no device to own, no radiation, and no skill to practise. It takes about a minute and works anywhere there is a phone. That removes the single biggest reason people stop tracking altogether: the method was too inconvenient to keep doing. A body fat number you never get around to measuring a second time is worth less than a shape trend you check every few weeks because it is genuinely easy to.

The honest limit: AI photo analysis reads visible shape and proportion, not internal composition. It cannot see bone density the way DEXA can, and it is not estimating lean mass from a physical signal the way bioimpedance is. What it is good at is exactly what a coach looking at progress photos over months is good at, noticing that your shoulders have broadened relative to your waist, or that your waist has narrowed relative to your hips, and putting a number and a trend line on an observation a mirror only gives you as a vague feeling. Pelaris renders that trend as a live 3D Body Blueprint, a figure shaped by your own proportions that you can rotate, play back across every check-in on a timeline, and compare side by side between any two dates. It is a visualisation of shape change, not a lab report, and it is built to be looked at often rather than booked occasionally. The ratios that sit alongside it, BMI, waist-to-hip and waist-to-height, are covered with their sources in what a body number can honestly tell you.

The five methods, side by side

Accuracy, from independent studies

What independent studies found when each method was compared with DEXA. The photo apps in the table are other apps, not Pelaris.

InBody, Evolt and other bioimpedance machines

Bioimpedance machines such as InBody and Evolt pass a small electrical current through the body and estimate body fat, muscle and water from how it travels. What a printout shows depends on the model, and higher models print more detail by body segment.

Studies have found that a meal or a drink changes the reading, so compare tests taken under the same conditions: the same machine, the same time of day and a similar hydration state. Against DEXA, a professional machine tends to show a steady offset rather than random error. In the InBody 770 study above, readings sat about 3.4 points under DEXA on average. A steady offset is something a trend can live with, which is why the change between your own tests is the more useful signal. Pelaris is not affiliated with InBody or Evolt.

InBody vs DEXA vs an AI body scan: which should you use?

It depends on the job you want the number to do.

• A measured checkpoint. DEXA. It is the reference method, with the caveat in the table above that two DXA machines can differ, so return to the same machine each time.
• Regular tracking where you train. Bioimpedance, such as an InBody machine or a smart scale, under the same conditions each time. Expect an offset against DEXA and use the trend.
• A quick estimate with no equipment. A tape measure and a published equation such as the Navy method, knowing from the table that it can differ from DEXA by a few points.
• Watching your shape change over weeks, from your phone. An AI photo scan, read as an estimate and a trend rather than a measurement. The photo apps in the studies above were other apps, and Pelaris has not been validated against DEXA, so it does not quote an accuracy figure.

Using them together in Pelaris

Take your scans, upload a DEXA or InBody result when you have one, and the measured result leads, with your photo estimate alongside it. Pelaris shows where each number came from.

Practical guidance: the best scan is the one you will repeat

Every method above can tell you something true. None of them can tell you anything useful as a single data point. The question worth asking before you pick one is not “which is most accurate,” it is “which one will I still be doing in three months, under close to the same conditions each time.”

A few things hold regardless of method:

• Keep conditions consistent. Same time of day, similar hydration, similar lighting if it is a photo or a scan, the same device if it is bioimpedance. The method matters less than the consistency of the method.
• Trend beats snapshot, every time. One reading is a photograph. A string of readings under similar conditions is a film, and the film is what tells you whether training is working. It is the same point we make about tracking training inputs instead of just outcomes: the single number lies by omission, the trend does not.
• Pair the scan with your training data. A body composition or shape trend means far more sitting next to what you actually did in the gym that month. Did the waist trend move during a block of consistent progressive overload, or during a month you barely trained? The scan tells you what changed. Your training log tells you why.
• Bring a DEXA or InBody result in. If you already have one, upload the PDF, a screenshot or a photo to Pelaris and it sits beside your photo scans as your measured result. Any other written record of your measurements works too, and you check every number before it is saved.
• Do not chase precision you will not act on. If a DEXA scan’s precise bone density reading will not change anything you do, a method you will actually repeat monthly beats one you book twice a year for a number you admire and then forget.

None of this says one method is right and the rest are wrong. The best method is the one that survives contact with your actual life. A DEXA scan you get once a year is worth having. A shape trend you check every few weeks because it takes a minute from your phone is worth more, not because it is more accurate, but because it is the one you will still be doing when it matters.

What this means for your training

• Body weight alone is too noisy to tell you whether training is working. Track composition or shape, not just the scale.
• DEXA is the accuracy benchmark, but it is clinic-bound and occasional. Treat it as a periodic checkpoint, not a weekly tool.
• Bioimpedance and 3D scanners are useful once you control the conditions, the same device, the same time, the same location, every time you use them.
• Tape and calipers are genuinely free and accurate, if you commit to consistent technique, which is the part most people skip.
• AI photo analysis trades clinical precision for frictionlessness: a photo or two, a minute, anywhere, which makes it the method most people will actually repeat. See how it works in Pelaris’s body scanning and 3D Body Blueprint.
• Whatever you choose, the trend is the signal. A single scan, however accurate, is a photograph. Keep taking it under similar conditions and it becomes a film.