People are increasingly turning to artificial intelligence with questions regarding symptoms, diagnoses, and medications. In Central Asia, this trend is already evident: 71% of respondents in Kazakhstan and 55% in Uzbekistan report using AI for health inquiries. Could such accessibility spark a surge in self-medication and exacerbate the crisis of antimicrobial resistance?
"Self-medicating with antibiotics based on information found online is one of the most perilous mistakes a patient can make. It poses a threat not only to the healthcare system, but far beyond it," says Orzu Rakhmonaliev, Director of the Republican Center for the Formation of a Healthy Lifestyle under the Ministry of Health and Social Protection of the Population of the Republic of Tajikistan.
According to him, a common error is using antibiotics to treat influenza, acute respiratory viral infections (ARVI), COVID-19, and acute viral gastroenteritis. Antibiotics have no effect on viruses, and their unwarranted use directly drives bacterial resistance to antimicrobial drugs.
For Tajikistan, the repercussions of this practice can be particularly severe. Rakhmonaliev draws attention to multidrug-resistant tuberculosis (MDR-TB). Indiscriminate use of certain antibiotics—specifically fluoroquinolones—for coughs or cold symptoms may temporarily mask clinical manifestations, delay timely TB diagnosis, and introduce significant health risks due to improper drug administration.
"When citizens purchase potent antibiotics over the counter at the first sign of a cough or cold, they may temporarily suppress symptoms. However, they lose critical time needed for an accurate diagnosis and appropriate clinical management," the expert explains.
Another major concern is the incorrect selection of drugs and dosages. Prescribing an antibiotic requires evaluating the patient’s age, body weight, renal and hepatic function, the nature and site of infection, and pathogen susceptibility.
Uncontrolled antibiotic use risks not only drug resistance, but also adverse drug reactions, disruption of the normal microbiome, allergic complications, and inflated treatment costs.
Orzu Rakhmonaliev also warns against the declining efficacy of first-line antibiotics—the most accessible and cost-effective treatments available. Consequently, clinicians are forced to rely on second-line or reserve antibiotics, which are often significantly more expensive, carry severe side effects, and are far less readily accessible.
Particularly dangerous is the tendency to discontinue therapy prematurely or modify the regimen once initial symptoms subside. This is especially hazardous in conditions like tuberculosis, where treatment demands strict adherence to an individualized regimen under close medical supervision.
AI Is Becoming the New Medical Interlocutor
But what happens when, instead of searching the web, a person turns directly to artificial intelligence?
"AI prompts and their influence on the decision to take antibiotics remain insufficiently studied," notes Rustam Gulov, an AI literacy specialist in the Republic of Tajikistan.
According to him, modern AI assistants are capable of recommending specific medications based on international clinical guidelines, usually appending a disclaimer about the necessity of consulting a doctor. However, the question remains whether users actually heed these warnings and consult professionals before initiating treatment.
"To varying degrees of success, ChatGPT, Claude, and Gemini are already replacing general practitioners for many people, suggesting various medications or outlining therapies," the expert says.
Rustam Gulov believes this trend will only intensify as medical AI services evolve. At the same time, he highlights another dimension of the issue: algorithms are not the only ones that make mistakes—physicians do, too. Looking ahead, he suggests that in certain medical tasks, AI may prove more accurate than humans. However, this still requires validation through clinical trials.
What Does Artificial Intelligence Itself Say?
While preparing this piece, we decided to put several questions to ChatGPT directly. It proved to be an exceptionally accessible respondent: requiring no prior appointment, waiting time, protracted approvals from management, or navigation around a packed schedule, it answered questions regarding medical consultations, self-medication, and antimicrobial resistance.
Can AI replace a doctor?
"Artificial intelligence is already capable of handling specific medical tasks: analyzing scans, assisting in interpreting laboratory results, detecting patterns, and assessing disease risks. However, it cannot fully replace a physician just yet. Medicine is not merely data analysis. A doctor takes into account medical history, individual patient characteristics, physical examination findings, and response to treatment."
Could the use of AI worsen the problem of antimicrobial resistance?
"Artificial intelligence can either mitigate antimicrobial resistance or exacerbate it. It all depends on how people utilize this technology."
Among the positive capabilities, ChatGPT highlights assisting physicians with the analysis of lab findings and antibiotic susceptibility data, early infection detection, epidemiological surveillance, and public education.
There is, however, a flip side. Individuals may describe symptoms to an AI and receive recommendations that they interpret as formal medical prescriptions. Erroneous information, self-selected antibiotic regimens, and the decision to forgo clinical care can introduce substantial risks.
What advice would you give to people turning to AI for self-medication?
"Do not use artificial intelligence to self-prescribe medications, especially antibiotics. Never start taking medications without a physician's consultation, and do not alter dosages or discontinue prescribed therapies based on an answer from a chatbot."
The AI itself acknowledges a fundamental limitation: it does not possess the full clinical picture of a patient and cannot serve as a substitute for a physical medical examination.
When Trust in the Algorithm Becomes Dangerous
The story of 31-year-old Bowen Wang, reported by The New York Times in October 2026, illustrates just how grave the consequences of over-relying on AI can be.
The woman suffered from epilepsy and was prescribed the anticonvulsant medication lamotrigine. According to a lawsuit filed by her parents against OpenAI, ChatGPT helped her design a tapering schedule to gradually reduce her dosage until completely discontinuing the drug. The lawsuit also alleges that the chatbot reinforced her decision not to disclose these changes in treatment to her doctors.
In February 2026, the woman passed away. The forensic medical examination determined the cause of death to be sudden unexpected death in epilepsy (SUDEP). Her parents believe that ChatGPT's guidance played a role in the tragedy and are seeking to hold the developer accountable.
It is important to emphasize that these are allegations outlined in a lawsuit, rather than a legally established causal link determined by a court.
While this case is not directly linked to antibiotics, it underscores a broader dilemma: a persuasive response generated by artificial intelligence can sway human decisions, even when it comes to life-sustaining medical care.
What Does the Research Show?
The trend of turning to digital tools for health information is corroborated by statistical data.
According to the international STADA Health Report 2026, conducted across 20 countries with 19,514 participants, 71% of respondents in Kazakhstan and 55% in Uzbekistan use artificial intelligence for health-related queries.
At the same time, 94% of respondents in Kazakhstan and 92% in Uzbekistan consider it acceptable to self-treat certain health issues. In Kazakhstan, 15% of respondents have experienced adverse effects from self-medication, compared to 22% in Uzbekistan.
However, widespread AI use does not mean people are ready to bypass doctors altogether. In Uzbekistan, 75% still prefer in-person clinical consultations, and 47% are unwilling to trust an AI diagnosis as much as a medical doctor's diagnosis.
A similar trend was identified in a study by The Health Management Academy in the United States, published in May 2026. The findings revealed that 57% of American adults initially turn to the internet or AI upon encountering unfamiliar symptoms before contacting a healthcare provider. Among them, 26% utilize AI tools such as ChatGPT, while 31% rely on standard search engines.
Earlier, in 2020, a study among 336 university students in the Indian state of Maharashtra revealed that approximately 71% of participants had attempted to self-diagnose based on online information at least once. Over 61% reported that searching for medical information heightened their anxiety.
These studies reflect the growing prevalence of digital self-diagnosis and self-treatment, though they do not conclusively prove that AI itself directly increases antibiotic consumption.
What Is Planned in Tajikistan?
Under the regional "One Health" program, the Ministry of Health and Social Protection of the Population of the Republic of Tajikistan is implementing measures outlined in the National Action Plan / Roadmap on Combating Antimicrobial Resistance for 2026–2030.
According to Orzu Rakhmonaliev, three key pillars of work are envisioned:
● Improving the Legal Framework: Reviewing and updating the regulatory framework governing antimicrobial circulation—from drug registration and importation to clinical prescription and utilization across human and veterinary medicine.
● Strengthening Monitoring and Prevention: Assessing epidemiological surveillance and infection control systems, expanding the monitoring of antibiotic consumption, and ensuring data exchange with the WHO Global Antimicrobial Resistance and Use Surveillance System (GLASS).
● Training and Public Education: Designing a curriculum on the rational use of antibiotics for healthcare workers and rolling out educational and awareness campaigns for the public.
There are also plans to formulate a national operational plan on infection prevention and control (IPC).
Between Technology and Responsibility
Tackling antimicrobial resistance requires the prudent use of antibiotics not only in human healthcare, but also in veterinary medicine, agriculture, and other relevant sectors under the "One Health" approach.
Artificial intelligence can serve as a valuable asset in this fight: helping analyze laboratory data, tracking pathogen resistance profiles, and assisting clinicians in optimizing treatment regimens. Yet, in the absence of medical oversight, it risks becoming another catalyst for unwarranted self-medication.
Research remains insufficient to assert definitively that AI usage directly fuels the rise of antimicrobial resistance. Nevertheless, the way people seek medical information is already undergoing a profound shift.
The challenge lies not merely in how intelligent AI becomes, but in how informed and literate its users are.
AI can help people understand an illness, but it must never independently prescribe treatment. In the containment of antimicrobial resistance, preserving the therapeutic efficacy of antibiotics goes hand in hand with upholding human accountability for personal healthcare choices.
Shoira Toirova — PR Specialist, "One Health" Program, shoira.tajikistan@gmail.com
This publication was developed by the Regional Environmental Centre for Central Asia (CAREC) within the framework of the regional "One Health" program in Central Asia. The program is implemented with financial support from the World Bank and the Pandemic Fund. The author bears sole responsibility for the content of this publication. The views expressed do not necessarily reflect the official position of the World Bank or the Pandemic Fund.
For further information:
Oxana Kravtsova — Regional Program Coordinator, "One Health" Program, CAREC, okravtsova@carececo.org