Opinion: The growing risk of AI-driven self-diagnosis in breast health

Artificial intelligence tools are rapidly becoming part of everyday life. From drafting emails to assisting with research, AI platforms are increasingly shaping how people access information. However, in health care, the rise of generative AI tools such as ChatGPT presents a new and potentially dangerous challenge.

More patients are now entering symptoms into AI chatbots in an attempt to diagnose themselves or determine whether a medical issue requires attention. While this behaviour may appear similar to traditional ‘Dr Google’ searches, generative AI carries a unique risk: it provides conversational responses that can sound authoritative, personalised and reassuring, even when the information is incomplete or misleading.

For clinicians, particularly those working in breast health, this trend is concerning. Early breast cancer diagnosis relies heavily on prompt clinical evaluation, imaging and appropriate referral pathways. When patients rely on AI-generated advice instead of seeking medical care, there is a real risk of delayed diagnosis.

Healthcare professionals need to be aware that AI is now influencing patient decision-making in ways that are not always visible during consultations.

A new form of digital reassurance

Patients have long searched the internet for health information, but generative AI tools change the nature of that interaction.

Instead of scanning multiple websites and forming their own interpretation, users can now ask a chatbot direct questions such as: “I have redness on my breast. Should I be worried?” or “Is a breast lump usually cancer?”

The response may appear thoughtful, structured and balanced, often including lists of possible causes and suggested actions. However, these tools cannot conduct a clinical examination, review imaging, understand an individual’s risk factors or interpret subtle contextual information.

In many cases, the responses are designed to provide general educational information rather than definitive guidance. Yet to a patient seeking reassurance, the answer can feel like a clinical opinion.

This is where the greatest risk lies.

False reassurance can lead to delayed presentation, particularly when symptoms are subtle or evolving.

Breast cancer does not always follow the textbook

Breast cancer remains the most common cancer affecting Australian women, with more than 21,000 diagnoses each year. Survival outcomes have improved significantly due to advances in screening, imaging, surgical techniques and systemic therapies.

However, early detection remains one of the most critical factors influencing prognosis.

The difficulty is that breast cancer does not always present with classic symptoms. While a palpable lump is the most recognised sign, many patients present with less obvious changes, including skin dimpling, nipple inversion, persistent redness, localised swelling or unusual discharge.

Inflammatory breast cancer, for example, may initially resemble an infection or skin irritation. These presentations require careful clinical assessment and imaging rather than observation alone.

When an AI tool suggests monitoring symptoms for several weeks before seeking medical attention, the consequences may be significant.

Even a short delay in investigation can allow aggressive cancers to progress.

AI cannot replace clinical assessment

Healthcare professionals understand that diagnosis is rarely based on symptoms alone. It requires integration of clinical examination, patient history, imaging and pathology.

AI tools cannot perform these functions.

They cannot palpate a breast lump, assess skin texture, evaluate asymmetry or detect subtle physical signs that experienced clinicians recognise during examination. They also cannot accurately interpret risk factors such as family history, genetic predisposition, hormonal exposure or previous imaging findings.

Even more importantly, they cannot recognise the clinical instinct that often guides experienced practitioners when something does not look right. Medicine is both a science and a practice refined by years of training and clinical exposure. AI platforms do not possess that judgement.

The clinical pathway exists for a reason

Australia has a well-established clinical pathway for investigating breast symptoms.

The process typically begins with a GP consultation, followed by imaging such as mammography or ultrasound when indicated. If imaging identifies a suspicious lesion, biopsy and specialist referral allow definitive diagnosis and management planning.

This structured pathway is designed to reduce risk and ensure that potential malignancies are investigated promptly and appropriately. When patients bypass these pathways in favour of digital reassurance, the system designed to protect them becomes less effective.

Healthcare professionals must therefore remain vigilant and proactive in reinforcing the importance of formal medical assessment.

The role of clinicians in the age of AI

Rather than ignoring the influence of AI, clinicians should acknowledge its growing presence and address it directly with patients.

Many individuals who consult AI platforms are not attempting to avoid health care; they are often seeking reassurance, convenience or preliminary information before deciding whether to book an appointment.

This behaviour presents an opportunity for clinicians to guide patients towards safer information practices. During consultations, healthcare professionals may consider asking patients whether they have searched online or used AI tools to interpret symptoms. Understanding this context can reveal misconceptions or false reassurance that may have influenced the patient’s decision-making.

Education remains one of the most powerful tools in addressing this issue. Patients should understand that while AI can provide general health information, it cannot provide diagnosis, clinical risk assessment or personalised medical advice.

Technology still has an important place in health care

None of this suggests that technology has no role in modern medicine. Digital tools, including AI, are already assisting clinicians in areas such as radiology, pathology analysis and administrative workflows. Its ability to synthesise data into information is unparalleled. When used appropriately and under clinical oversight, these technologies can improve efficiency and enhance patient care. AI may also help disseminate public health information, increase awareness of screening programs and improve accessibility to educational resources.

However, there is a fundamental distinction between technology that supports clinicians and technology that attempts to replace clinical decision-making.

Breast health is an area where nuance, experience and careful assessment are essential. The stakes are simply too high for unsupervised digital interpretation.

A message for healthcare professionals

As AI platforms become more accessible, healthcare professionals must anticipate that patients will continue to use them.

Rather than viewing this trend purely as a technological issue, it should be recognised as a new dimension of patient behaviour. Clinicians should actively reinforce several key messages:

• AI tools cannot perform clinical examinations or diagnostic imaging.
• Symptoms involving the breast should always be assessed by a qualified medical professional.
• Any persistent breast change, lump, discharge, skin change or unexplained pain warrants medical evaluation.
• Early detection remains the most effective strategy for improving outcomes. Encouraging patients to seek timely medical advice remains one of the most powerful interventions healthcare professionals can offer.

Protecting patients through clarity and communication

Artificial intelligence will continue to evolve and influence how patients interact with health information. For healthcare professionals, the challenge is not to compete with technology but to ensure patients understand its limitations. When it comes to breast health, the message must remain clear.

A chatbot cannot examine a patient, it cannot order imaging and it cannot diagnose cancer. Only trained clinicians working within established medical pathways can provide that level of care.

In an era of rapidly advancing technology, one principle remains unchanged: if a patient notices a breast change or symptom, the safest course of action is to seek medical assessment without delay. That advice is far more reliable than anything an algorithm can provide.

*Associate Professor Sanjay Warrier is a past president and current committee member of Breast Surgeons of Australia and New Zealand (BreastSurgANZ). His views are his own, not BreastSurgANZ. Warrier’s surgery is located at the Chris O’Brien Lifehouse and he also has clinics at Oxford Street, Bondi Junction and Macquarie Street, Sydney. He is published in numerous peer-reviewed journals and won the Royal Prince Alfred Hospital’s Patron’s Prize for best scientific research. He recently received the India Australia Business and Community Alliance (IABCA) Australia India Science, Research and Development Award.

Top image caption: Associate Professor Sanjay Warrier in a consultation. Image: Supplied