A new study led by Stanford researchers and co-authored by UC Berkeley professor Ziad Obermeyer found that 98.8% of Alodokter AI’s real-world telemedicine consultations were clinically safe, based on independent physicians’ end-to-end review of the consultations.
Alodokter AI is developed by Alodokter, a leading digital health company based in Indonesia with 20 million monthly active users. The AI is deployed as an end-to-end support tool for doctors during chat-based teleconsultations.
In this study, independent doctors evaluated 402 consultations led by Alodokter AI across 14 safety-based criteria, including history-taking, red-flag detection, diagnosis, medication and testing, triage, referral, safety-netting and hallucination control. 397 consultations were classified clinically safe.
Based on available research, Alodokter AI is one of the first deployed LLM-based telemedicine systems independently evaluated for clinical safety across the full consultation workflow in real-world telemedicine consultations.
“Medical AI cannot be judged only by how well it answers curated questions. Real-world consultations are complex and unpredictable. By evaluating Alodokter AI across complete, real-world telemedicine consultations, this study brings clinical AI evaluation closer to the conditions that truly matter for patient care,” said Ziad Obermeyer, MD, an associate professor at UC Berkeley and one of TIME’s 100 most influential people in AI.
Deployed in the Alodokter app since January 2026, Alodokter AI works in the background to support doctors throughout each teleconsultation. Doctors remain in control. A licensed physician reviews the AI’s recommendations, can modify or reject them, and signs off the final care. Their feedback on patient cases is used to continuously improve the system.
“LLMs have opened extraordinary possibilities in medicine. Turning those capabilities into reliable support for medical consultations requires deep clinical experience and rigorous evaluation. The stakes are much higher in a medical consultation than in a broad generic medical answer. Our technology is built to manage the complexity and safety-challenges of real-world teleconsultations,” said Dr. Louise Hewitt, Alodokter Chief Medical Officer.
Alodokter AI is now used by general practitioners across its telemedicine platform, with adoption approaching 100%. Internal data show that 98% of consultations with general practitioners supported by Alodokter AI received positive patient satisfaction ratings. This rapid doctor and patient adoption reflects the value of an AI built for the complexity of telemedicine consultations.
“We built Alodokter AI on more than 10 years of experience managing tens of millions of teleconsultations. We see a major opportunity to help doctors serve more patients while maintaining the quality and personal attention that good care requires. By supporting the full consultation, we aim to make high-quality medical expertise more accessible and enable telemedicine to scale,” said Nathanael Faibis, Alodokter CEO.
The study put that ambition to the test using real patient consultations, with a sample deliberately enriched for emergency, referral and other higher-acuity cases. This made it a demanding assessment of clinical safety under the conditions doctors encounter in everyday practice.
“Clinical AI should be judged in the healthcare systems where it will actually be used. This study combines real-world patient consultations, independent medical review and physician oversight, providing the type of evidence health systems need to evaluate clinical AI in real-world use and support its wider adoption,” said Wendy Yin and Natalia Khoudian, doctoral researchers at Stanford University.
Alodokter is now working with the same research team on a randomized controlled trial to evaluate the impact of Alodokter AI on clinical quality and doctor productivity. The trial is designed to provide the next level of evidence needed to scale Alodokter AI safely.
Read the study:
https://kingcenter.stanford.edu/publications/health/assessing-clinical-safety-ai-assistant-indonesian-chat-based-telemedicine