Digital biomarker research has generated thousands of publications demonstrating associations between sensor-derived measures and clinical conditions, yet clinical adoption remains negligible. We identify a foundational problem: the field lacks consensus on what constitutes a digital biomarker, applying identical terminology to direct physiological measurement (continuous glucose monitoring), algorithmic prediction of biological substrates (voice analysis for dopaminergic function), and purely behavioral correlates (GPS mobility and depression scores). This terminological ambiguity obscures validation requirements and prevents evidence synthesis. We argue that the “bio-” in “digital biomarker” refers to a property of the measurement itself—the marker must be derived from biology, not merely predictive of biological or clinical outcomes. Under this restrictive definition, behavioral correlates without demonstrated biological grounding, however statistically robust or clinically useful, should be designated as digital phenotypes or digital health indicators rather than biomarkers. This distinction clarifies validation pathways: biologically derived markers require technical accuracy validation against established biological reference standards; candidate biomarkers inferred from behavior require biological criterion validation before that status is claimed; behavioral indicators require outcome prediction validation. We demonstrate how this framework resolves current validation confusion and accelerates translation by aligning evidence standards with measurement types.
