India has never been short of ways to obtain a medication without seeing a doctor. Artificial Intelligence or AI may simply add a new one. On October 5, Utah allowed Nolla Health to begin a supervised pilot in which its AI system assesses photographs of mild-to-moderate acne and can generate first-time prescriptions or refills for a limited set of topical medicines. The pilot applies only to adults in Utah, excludes severe acne and several higher-risk groups, and does not cover oral medicines or isotretinoin. Two Utah-licensed physicians must review every prescription before it reaches a pharmacy, and any later move to retrospective review or sample-based checks will require separate regulatory approval.
Utah is testing whether human review can be reduced gradually without losing control over prescribing. In its separate August pilot for maintenance medicines, clinicians review the first 250 refills before they are issued and the next 1,000 after the event. If regulators allow the pilot to proceed further, at least 5% of prescriptions in each medication class will still be checked. Any case that raises a safety concern or lacks required monitoring is removed from the automated pathway and returned to a clinician. Lower-risk cases receive less supervision only after the system has been observed in use, while uncertain cases still go back to a human reviewer. This is a more cautious form of automation than simply using software to make up for shortages of medical staff. In India, it should be judged against the care people actually receive when they self-medicate or obtain prescription medicines informally from pharmacies, rather than against an ideal consultation that may never take place.
On paper, India already keeps a fairly tight grip on prescription medicines. Schedule H and H1 drugs are meant to be dispensed only on a doctor’s prescription. At the pharmacy counter, practice is often more permissive. A recent systematic review of 66 Indian studies, covering nearly 30,000 participants, put the pooled prevalence of self-medication at 64.4%, though the underlying studies differed widely. People often relied on pharmacists, previous prescriptions or their own experience when deciding what to take. Antibiotics appeared often enough in the evidence to make the gap between regulation and everyday practice hard to dismiss.
{{/usCountry}}On paper, India already keeps a fairly tight grip on prescription medicines. Schedule H and H1 drugs are meant to be dispensed only on a doctor’s prescription. At the pharmacy counter, practice is often more permissive. A recent systematic review of 66 Indian studies, covering nearly 30,000 participants, put the pooled prevalence of self-medication at 64.4%, though the underlying studies differed widely. People often relied on pharmacists, previous prescriptions or their own experience when deciding what to take. Antibiotics appeared often enough in the evidence to make the gap between regulation and everyday practice hard to dismiss.
{{/usCountry}}Studies of Indian pharmacies have repeatedly found prescription-only antimicrobials being sold without prescriptions despite their Schedule H or H1 status. This gives the debate over AI prescribing a different texture in India. The obvious concern is that an understaffed health system might use software as a cheap substitute for doctors. But many patients are already making treatment decisions without the sort of clinical consultation that regulation assumes. Any assessment of AI prescribing in India, therefore, has to begin with the system people actually use, not the one described in the rulebook.
Indians already make a large share of everyday health decisions without seeing a doctor. The country’s OTC market was worth about ₹47,000 crore in 2024, and EY-Parthenon expects it to reach nearly ₹98,000 crore by 2030. Its survey found that pharmacists were the preferred source for 80% of pain-relief purchases and 67% of purchases related to chronic health conditions. More than four-fifths of respondents sought professional or expert advice when buying an OTC product for the first time, although familiarity and brand trust played a larger role in repeat purchases. The neighbourhood pharmacist still occupies an unusual place in Indian health care, somewhere between shopkeeper and first point of advice.
AI is now becoming another source of health advice. In a 2025 BCG survey of more than 13,000 internet-connected adults across 15 countries, nearly 60% said they had used generative AI for some aspect of their health, with particularly high use in countries including India. Most of this activity still falls short of formal diagnosis or prescribing, but the boundary is already becoming harder to police. A patient may ask a chatbot about a rash and then approach a pharmacy for the medicine it suggests. The prescription may still belong legally to the doctor, but the treatment decision may have started elsewhere.
If Utah eventually allows the acne system to operate with only retrospective checks, the decision will not stay confined to one pilot. Future applicants will be able to point to an approved example in which prospective physician review was reduced after a period of supervised use. The slippery slope lies there. A company seeking approval for another skin condition, for instance, would no longer be arguing from first principles; it would be asking the regulator to move a boundary that has already moved once.
India has an additional complication. Schedule H and H1 controls already weaken at the point of sale, while patients increasingly reach pharmacies after looking up symptoms or treatments online. Formal approval for AI prescribing would give such advice a different status. Once software is permitted to issue a prescription in one setting, consumers and pharmacies are likely to treat its recommendations with greater authority elsewhere too. Regulators in India would have to resist treating one successful use case as evidence for the next. Approval for acne should remain approval for acne. A proposal involving eczema or another apparently similar condition should return with evidence specific to that use case and to Indian prescribing conditions. Otherwise, one exception can become the basis for another.
Britain offered a useful contrast on October 6, when the government accepted all 44 recommendations of its National Commission on AI in Healthcare and moved ahead with another phase of the MHRA’s AI Airlock. The commission argued for continued scrutiny after approval because performance can change once software is updated or used across patient groups that were poorly represented during testing. India’s version of this debate will probably be decided across pharmacy counters. Once an AI recommendation can legally become a prescription, a habit that was informal acquires regulatory sanction. Approval for acne should end with acne.
(The views expressed are personal)
This article is authored by Dr KS Uplabdh Gopal, health professional and Dr K Madan Gopal, professor, community medicine and former senior consultant, health, NITI Aayog.