AI in Healthcare: High Adoption Among Indian Clinicians, Yet Reliance on Generic Tools Raises Patient Safety Concerns, ETHealthworld

Artificial intelligence (AI) is quietly becoming a part of clinical practice in India, helping doctors search medical literature, draft clinical notes, educate patients and even support bedside decision-making.

However, as this adoption accelerates, healthcare experts are raising a concern: Many clinicians are increasingly relying on generic AI chatbots that were never designed for this field.

A new global survey by Elsevier, a scientific publisher, has found that while AI adoption among Indian clinicians continues to rise, a majority of those using it frequently rely on general-purpose large language models (LLMs) rather than clinical-specific AI platforms built on verified medical literature.

According to the Clinician of the Future 2026 report, 48per cent of Indian doctors and clinicians surveyed use AI for work. Of them, 54per cent frequently or always use general-purpose AI tools, compared with only 26per cent who regularly use clinical-specific platforms. Similar trends are seen among nurses, with 71per cent of AI users relying on general AI tools against 46per cent using specific systems.

Healthcare experts say though this trend reflects a growing comfort with AI, it exposes a patient-safety gap.

“Many clinicians are turning to generic LLMs such as ChatGPT or Claude during clinical workflow. These tools were never designed for medical decision-making. They can produce very confident answers, but a confident answer is not necessarily the correct one,” Dr Rahul Goyal, lead clinical executive at Elsevier and a physician who has worked across India, UK, Europe and West Asia, told TOI.

He said unlike consumer chatbots that may draw information from across the internet, including non-peer-reviewed sources, clinical AI tools like ClinicalKey AI are trained on curated medical textbooks, journals and clinical guidelines. If sufficient evidence is unavailable, they are expected to indicate that rather than generate a plausible response.

The distinction, however, becomes critical when a doctor must make a diagnosis or decision on treatment while a patient is sitting in front of them, Goyal said. “There is virtually no room for error in these situations. If a clinician relies on a generic AI tool and receives an incorrect recommendation, the legal responsibility still rests with the doctor. AI does not absorb that liability,” he said, while describing this as “liability sink” — a concept under which the clinician remains accountable even if the advice originated from an AI system is wrong and remains uncorrected.

The findings come at a time when AI is rapidly entering hospitals worldwide. Elsevier’s survey of 2,757 clinicians from 118 countries found that 49per cent of clinicians globally use AI at work, while 80per cent believe AI will become a critical assistant in healthcare within the next decade.

In India, where the maximum number of clinicians were surveyed — 426 doctors and 112 paramedics — general AI tools are primarily used for medical research (57per cent ), professional education (57per cent ), identifying drug information (50per cent ), patient education (46per cent ), clinical decision support (40per cent ) and clinical documentation (40per cent ).

Despite the concerns, the study does not see AI replacing clinicians, and believes that properly designed clinical AI could improve the consistency of care, particularly in countries, such as India where access to specialists remains uneven.

“In a country as large as India, technology can help narrow the variation in treatment between metropolitan hospitals and those in smaller towns. Clinical AI can make medical care more uniform by supporting doctors with evidence-based recommendations instead of individual practice patterns,” Goyal said.

He added that AI should be viewed as a clinical assistant, not a replacement for medical judgement. “Doctors who use AI responsibly will be able to care for more patients and make better-informed decisions than those who don’t. But AI is a tool, not a decision-maker. The clinician must always remain in the loop.”

The report also suggests that Indian clinicians are thinking critically about trustworthy AI. Doctors said they would have greater confidence in AI systems that are easy to use (68per cent ), validated by independent clinical experts (67per cent ), transparent in citing references (64per cent ), trained to avoid harm (63per cent ), and draw information from multiple high-quality sources (63per cent ).

Senior clinicians said AI can enhance healthcare but should never replace a doctor’s judgement. According to Dr G C Khilnani, chairman, PSRI Institute of Pulmonary, Critical Care and Sleep Medicine, and former professor AIIMS, AI can help analyse investigations and improve efficiency. However, “medicine is both science and art, and AI cannot replicate a physician’s clinical acumen or recognise every atypical presentation. The responsibility for patient care will always rest with the doctor,” he said.

Dr Latika Bhalla, senior adolescent paediatrician at Sir Ganga Ram Hospital, said AI is useful for generating differential diagnoses, maintaining records and reducing paperwork, but “diagnosis and treatment must always be guided by clinical experience”.

Dr (Maj.) Rajesh Bhardwaj, a specialist at MedFirst ENT Centre and professor emeritus, Tirthankar Mahavir University, said while AI is an excellent assistant, doctors should rely on specialised medical AI, verify AI-generated information against established clinical guidelines, and ensure that patient confidentiality is never compromised.