AI Diagnostics in Physical Therapy: Attitudes, Barriers, and Clinical Implications

Introduction: AI in a hands-on field

Artificial intelligence (AI) is transforming many corners of healthcare, yet physical therapy has been slower to adopt AI-powered diagnostics compared with some medical specialties. Physical therapists (PTs) traditionally rely on hands-on assessments, patient history, and nuanced clinical reasoning to determine diagnoses, prognoses, and treatment plans. As AI tools mature—from motion analysis analytics to predictive risk models—therapists face a pivotal question: how do clinicians feel about integrating AI diagnostics into everyday practice, and what factors will decide whether such tools improve or hinder patient care?

What therapists think: current attitudes toward AI diagnostics

Across surveys and qualitative studies, PTs express a mix of cautious optimism and practical skepticism. Many see AI as a potential augmenter to clinical judgment, offering consistent data, objective measurements, and faster triage. Others worry about overreliance on algorithmic recommendations, loss of tactile expertise, and the risk that AI could deprioritize individualized, patient-centered care. A common thread is the belief that AI should function as a decision-support tool rather than a replacement for hands-on assessment and clinical acumen.

Barriers: what stands in the way of adoption

Several barriers hinder the uptake of AI diagnostics in physical therapy:

• <strong Trust and interpretability: PTs want transparent models that explain how a recommendation was reached and what data influenced it.
• <strong Data quality and integration: Inconsistent documentation, incompatible systems, and fragmented patient data impede reliable AI performance.
• <strong Clinical relevance: There is concern that AI metrics may not capture essential manual assessments or the nuances of musculoskeletal conditions.
• <strong Workflow disruption: Tools that complicate visits or add steps without clear value are unlikely to be adopted.
• <strong Legal and ethical considerations: Liability, patient privacy, and informed consent around AI-derived recommendations are ongoing debates.
• <strong Training gaps: Limited exposure to AI in PT education means many clinicians lack confidence in interpreting outputs or integrating them into plans of care.

Enablers: what could accelerate responsible adoption

When designed with input from PTs, AI diagnostics can become valuable allies. Key enablers include:

• <strong Clinically relevant outputs: AI provides actionable insights that align with PT assessment domains (movement quality, functional tests, bending and gait analyses).
• <strong Interoperable data systems: Seamless data sharing between EHRs, wearable devices, and clinic software reduces manual entry and errors.
• <strong User-centered design: Intuitive interfaces, minimal disruption to workflow, and on-demand explanations boost trust and usability.
• <strong Education and evidence:case-based demonstrations, peer-reviewed research, and continuing education improve comfort with AI tools.
• <strong Clear governance: Guidelines on accountability, safety checks, and patient consent help navigate ethical concerns.

Clinical implications: how AI could reshape PT practice

Appropriate AI diagnostics have the potential to augment professional decision-making without erasing the PT’s central role. Possible implications include:

• <strong Enhanced assessment: Quantified movement patterns and risk stratification can inform targeted interventions and progression tracking.
• <strong Personalized care: AI may help tailor treatments by identifying which modalities yield the best response for a given patient profile.
• <strong Proactive prevention: Early detection of functional decline could trigger timely referrals or preventive strategies.
• <strong Resource optimization: Clinicians can allocate time more efficiently by focusing attention where AI flags higher risk or complexity.

Recommendations for stakeholders

For successful integration, stakeholders should prioritize co-design with PTs, rigorous clinical validation, and phased implementation that preserves the clinician-patient relationship. Academic programs and professional bodies can provide standardized curricula and guidelines, while technology vendors should emphasize transparency, data stewardship, and interoperability. Ultimately, AI diagnostics should extend the PT’s reach—supporting, not replacing, the therapeutic alliance that drives functional recovery.

Conclusion: toward wise, patient-centered AI use in PT

Attitudes toward AI diagnostics among physical therapists are diverse but generally moving toward cautious adoption anchored in clinical judgment and patient safety. By addressing barriers like interpretability and workflow integration while leveraging enablers such as clinically meaningful outputs and education, AI can become a constructive partner in physical therapy—enhancing assessment, personalizing care, and preserving the core human elements of therapy.