Australian Dermatologists’ Perspectives on AI Integration: Trust, Benefits, and Barriers in Clinical Practice

Highlight

– Nearly half of Australian dermatologists have experimented with AI, but regular clinical use remains low (12%).
– Trust issues, especially regarding diagnostic accuracy for skin cancer, present a major barrier to AI adoption.
– Clinicians desire transparency about AI datasets, limitations, and benefits to build trust.
– While few fear AI will replace dermatologists, many anticipate AI will perform key dermatological tasks.

Study Background and Disease Burden

Artificial Intelligence (AI) is increasingly transforming healthcare, promising improvements in diagnostic efficiency and patient access. Dermatology is a specialty amenable to AI applications, particularly in skin cancer diagnosis—a major public health concern frequently requiring timely expert evaluation of lesions. However, the successful adoption of AI technology depends on clinician acceptance and trust. Understanding attitudes amongst dermatologists is critical given their pivotal role in integrating AI into routine clinical workflows.

In Australia, where skin cancer rates are among the highest worldwide, advancing tools that aid early and accurate diagnosis is imperative. Yet, the potential risks such as AI inaccuracy, data privacy concerns, and ethical implications remain salient. This study aimed to explore how Australian dermatologists perceive the role of AI in their practice—including perceived benefits, barriers, and future expectations—providing insights to guide effective implementation.

Study Design

This research utilized an anonymous, cross-sectional online survey targeted at Fellows and Trainees of the Australasian College of Dermatologists (ACD). Invitations were distributed broadly, resulting in 122 completed responses, a 16.2% response rate. The survey comprised quantitative and qualitative items assessing AI usage patterns, trust levels, perceived accuracy, benefits and risks, and projected impact on the dermatology workforce.

Endpoints focused on the proportion of dermatologists using AI clinically or administratively, trust in AI for skin cancer diagnosis, requisite accuracy standards, and views on the future integration of AI into dermatology practice.

Key Findings

Usage Patterns: Among respondents, 44% reported having used AI tools in their dermatology practice; however, routine usage was limited to 12% for clinical purposes and 17% for administrative support. This reflects cautious adoption rather than widespread integration.

Trust and Accuracy Concerns: Trust emerged as the greatest obstacle to AI utilization, with 69% either unwilling or uncertain about relying on AI to support skin cancer diagnosis. More than half (52%) expressed that AI should be at least as accurate as the best dermatologist to be clinically acceptable.

Participants emphasized the need for transparent data regarding the AI algorithms’ training datasets, clear delineation of AI limitations, and explicit communication about the tool’s intended function to foster trust.

Perceived Benefits: Dermatologists recognized that AI could alleviate monotonous, repetitive tasks, potentially streamlining workflow and reducing clinician burnout. Improved patient access, particularly in underserved or remote areas, was viewed as a key advantage.

Perceived Risks: Accuracy concerns dominate risk perception, underscoring apprehension about misdiagnosis or missed diagnoses if AI is over-relied upon. Additionally, divestment or excessive control of AI technologies by technology companies raised apprehension about commercial interests overriding clinical priorities.

Future Outlook: While only 10% feared complete replacement by AI, nearly half (47%) anticipated that key aspects of dermatology work might be delegated to AI systems. This indicates recognition of AI as an adjunct rather than a substitute.

Expert Commentary

The survey highlights a critical juncture in dermatology’s digital evolution. The reticence to fully embrace diagnostic AI tools without robust evidence of accuracy and transparent validation underscores the profession’s commitment to patient safety. Similar trends have been documented internationally, where clinician skepticism centers on “black-box” algorithms and potential biases within training datasets (Esteva et al., 2017; Tschandl et al., 2020).

The expressed preference for AI performance at least matching the best clinicians aligns with guidelines emphasizing that AI should augment, not replace, human judgment (American Academy of Dermatology, 2022).

These findings suggest that professional bodies like the ACD can play a pivotal role by providing educational resources, setting validation standards, and facilitating clinician feedback in AI development. Collaborative efforts with technology developers are essential to align AI tools with clinical workflows and ethical standards.

Conclusion

Australian dermatologists stand at the forefront of integrating AI into clinical practice but predominantly remain cautious, particularly concerning diagnostic accuracy for skin cancer. Trust barriers and concerns about data transparency must be addressed to foster wider adoption. The perceived potential benefits of workflow efficiency and enhanced patient access are encouraging, provided AI tools are rigorously validated and deployed under clinician oversight.

Professional dermatology organizations have an opportunity and responsibility to guide clinicians through the evolving AI landscape by promoting informed discourse, evidence-based adoption practices, and safeguarding patient-centered care. Further research should explore longitudinal changes in acceptance, real-world AI effectiveness, and strategies to optimize human-AI collaboration in dermatology.

References

– Partridge B, Janda M, Gillespie N, Silva CV, Arnold C, Abbott L, Caccetta T, Soyer HP. Attitudes Towards the Use of Artificial Intelligence in Dermatology: A Survey of Australian Dermatologists. Australas J Dermatol. 2025 Aug;66(5):e279-e286. doi: 10.1111/ajd.14524.
– Esteva A, Kuprel B, Novoa RA, Ko J, Swetter SM, Blau HM, Thrun S. Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017 Feb 2;542(7639):115-118.
– Tschandl P, Codella N, Akay BN, Argenziano G, Braun RP, Cabo H, et al. Comparison of the accuracy of human readers versus machine-learning algorithms for pigmented skin lesion classification: an open, web-based, international, diagnostic study. Lancet Oncol. 2020 Jul;21(7):938-947.
– American Academy of Dermatology. Position statement on artificial intelligence in dermatology. 2022.