Tests that report a biological age produce a single number that is often lower or higher than the calendar age. That number comes out of a statistical model, and understanding the model explains what the result can and cannot mean.
The measurement underneath the number
Most consumer biological age tests read methylation, a pattern of small chemical marks attached to DNA that influence whether genes are switched on.
Methylation at particular sites shifts in fairly consistent directions as people get older, which makes those sites usable as an ageing signal.
A saliva or blood sample is processed to give a methylation value at each of hundreds of thousands of sites, and that raw profile is the actual measurement.
A model converts the profile into an age
The profile means nothing on its own. It is fed into a model that was trained on samples from people whose calendar ages were already known.
Training finds the combination of sites that best predicts age in that group, and the resulting formula is applied to new samples.
The output is therefore a prediction of what age someone with this methylation profile would typically be, not a direct reading of any biological process.
Why different tests disagree
Different clocks were trained on different populations and against different targets — some against calendar age, some against mortality risk or measures of physical function.
Two clocks run on the same sample can return results several years apart because they are answering slightly different questions.
Comparing a result from one provider against a result from another has little meaning unless both were built the same way, which is rarely disclosed in detail.
Repeat testing carries substantial noise
Sample handling, the proportion of different cell types in a blood draw, and processing batch all affect methylation readings.
The result is that repeating a test on the same person within a short period can shift the number by a year or more without anything having changed biologically.
A single small movement between two tests is therefore weak evidence of anything, which is a limitation the format of the report tends to hide.
What the number is useful for
Across large groups, clocks trained against health outcomes do carry information: populations scoring older tend to experience more disease over follow-up.
A population-level association is a much weaker statement than a personal prediction, and the tests are not diagnostic instruments.
Anyone using a result to make decisions about symptoms or treatment needs a clinician working from clinical measures, because a biological age figure does not identify what is happening in a specific body.