How Accurate Is AI at Diagnosing Illness? A GP Explains the Risks

Across the Western Cape, more patients are turning to artificial intelligence for quick medical answers. With technology readily at our fingertips, many people now research symptoms online or consult AI tools before visiting a doctor. But the real question many are asking is this: How Accurate Is AI at Diagnosing Illness, and can it be trusted with something as important as your health?

Dr Theresa Jordaan, a GP with more than 30 years of experience, believes AI has enormous potential—yet it must be used responsibly and always alongside a qualified medical professional. Even the Western Cape Government is exploring AI to modernise healthcare services, but these tools are designed to support clinicians, not replace them.

Can AI Diagnose Diseases Safely?

AI can be incredibly helpful. It can summarise research, explain symptoms, and help patients feel more informed before stepping into a consultation. Many patients at our practice bring AI‑generated information to their appointments, and Dr Jordaan welcomes this. It strengthens the shared‑care model and helps patients feel more involved in their health decisions.

But despite these benefits, AI is not a doctor. Large language models can “hallucinate”—a term used when AI generates medical facts, diagnoses, or medication dosages that sound authoritative but are completely incorrect. A 2025 analysis published by Patient Safety & Quality Healthcare found that even advanced medical AI systems can produce dangerous or implausible answers, including incorrect diagnoses and fabricated clinical details.Reference: AI in Healthcare: Addressing the Reality of Hallucinations

This is why AI should never be used as a standalone diagnostic tool. You can explore our team of qualified doctors here.

Why AI Gets Some Symptoms Right and Others Dangerously Wrong

AI models are trained on vast datasets, but not all datasets represent everyone. Diet, stress levels, inflammation, genetics, and lifestyle vary widely across individuals and communities. In the Western Cape, these differences are even more pronounced due to our diverse population and environmental factors.

AI cannot account for:

• Your personal medical history

• Your physical appearance when healthy

• Subtle changes in your voice, posture, or behaviour

• Cultural or lifestyle nuances that influence health

  
  

Dr Jordaan often recognises when a patient is unwell simply because she has seen them when they were healthy. AI does not have this privilege. It cannot listen to your chest, examine your skin, or detect early warning signs that only a trained clinician would notice.

For safe, personalised care, a physical examination remains essential. Learn more about our GP services here.

How Accurate Is AI at Diagnosing Illness? What Patients Should Understand

AI can appear highly knowledgeable, especially when it provides quick, confident answers. But accuracy varies widely, and AI often struggles to distinguish between conditions with overlapping symptoms. While it may correctly identify common patterns, it cannot evaluate physical signs, emotional cues, or subtle clinical details that a GP immediately notices. This is why AI should be viewed as a supportive tool—not a diagnostic authority.

How Accurate Is AI at Identifying Symptoms? What Patients Often Miss

AI can be surprisingly accurate at identifying common symptoms or offering general explanations. But accuracy is not the same as diagnosis. AI may correctly describe a symptom—such as chest tightness or fatigue—yet completely misinterpret its cause.

For example:

• Chest tightness could be anxiety… or a heart attack.

• Fatigue could be stress… or anaemia, thyroid disease, or infection.

• A rash could be an allergy… or an autoimmune condition.

AI cannot differentiate between these possibilities without examining you. This is where the danger lies: AI sounds confident, even when it is wrong.

Can AI Get Your Diagnosis Wrong?

Yes, and this is one of the biggest risks. AI tools can produce:

• Incorrect diagnoses

• Wrong medication dosages

• Misleading explanations

• Fabricated medical facts

These errors can be subtle, persuasive, and difficult for patients to detect. A doctor, however, will immediately recognise when something is inaccurate or unsafe.

Who Is Responsible When AI Gets a Diagnosis Wrong?

This is one of the most important questions patients are asking: Who is accountable if AI harms you?

At present, there is no clear legal framework in South Africa—or globally—that defines liability for AI‑generated medical advice. If an AI tool gives you incorrect information that leads to harm, there is no established system for accountability.

When patients and doctors review AI information together, they can identify what is safe, what is inaccurate, and what requires further investigation. This shared decision‑making protects the patient and ensures that AI becomes a helpful tool rather than a risky shortcut.

If you have AI‑generated medical questions you’d like to discuss, you can contact the practice here.

Is AI Good Enough to Diagnose You? Why Human Doctors Still Matter

Artificial intelligence is the future of medicine, and it is an exciting tool that both patients and healthcare providers should embrace. But AI cannot replace the clinical judgement, intuition, and contextual understanding of a trained GP.

The Western Cape’s own investment in AI‑driven healthcare modernisation shows how powerful these tools can be—but importantly, the province is using AI to enhance clinical workflows, not to replace the essential judgement of trained medical professionals.

Dr Jordaan encourages patients to bring AI findings to their consultations. Together, you can decide what is safe, what is relevant, and what requires further testing or examination.

---

About Dr Theresa Jordaan

Dr Theresa Jordaan is the founder of the practice and a compassionate GP with more than 30 years of experience. She is known for transforming healthcare delivery and leading child‑safeguarding initiatives. Her approach blends deep clinical expertise with a warm, patient‑centred philosophy.