When a young woman in Cameroon noticed an unusual change in her vaginal discharge, she wanted to know whether something was wrong. Rather than visit a health facility, she turned to ChatGPT.
The discharge was thick and brown. Concerned, she took a photograph and uploaded it to the artificial intelligence (AI) chatbot, asking whether it was normal.
The response reassured her, advising her to monitor the symptoms and seek medical attention if they persisted, developed an unusual smell or caused further concern.
“I thought something was wrong with me, but after asking AI, I felt relieved because I learnt that it was normal,” she recalled.
Within a few days, the discharge had stopped. She believed the chatbot had helped her understand her body and eased her anxiety.
But the disappearance of a symptom does not establish what caused it, and a chatbot’s reassurance is not a medical diagnosis. Without a clinical assessment, it is impossible to know whether the response correctly explained what happened.
For the young woman, however, she had obtained an answer without discussing an intimate concern with another person or immediately paying for a consultation.
“Medical checkups are more expensive than AI,” she said.
Her experience captures both the promise and the risks of using AI for personal healthcare. Chatbots can offer a private first stop for people seeking quick answers about their bodies. But when that first stop becomes the final authority, general information can be mistaken for a diagnosis.
A private answer to an intimate question
For many young people, questions about reproductive and sexual health can be difficult to raise. Embarrassment, fear of judgement, concerns about confidentiality, consultation costs and limited access to health professionals may influence where they seek answers.
A chatbot offers a different experience. Users can ask questions at any hour, describe symptoms in their own words and receive responses within seconds. Some platforms can also interpret images and ask follow-up questions about symptoms and personal circumstances.
Aurelia, a youth leader who has used ChatGPT for health-related concerns, says she has turned to it when experiencing stomach problems and looking for possible remedies to try at home.
She values the options it provides, particularly when they serve as a starting point for further investigation.
“It gives me more like things that I could do, that I could check out and verify with a health practitioner that will give you real advice and medications to take,” she explained.
Aurelia says the chatbot sometimes asks follow-up questions about what she has eaten, whether she is menstruating, allergies and aspects of her lifestyle before offering suggestions.
Although she describes these exchanges as AI diagnosing her, asking questions and suggesting possible explanations are not the same as establishing a medical diagnosis.
A detailed response can sound authoritative even when the information available is incomplete. For someone worried about a symptom, the difference between “this could be one explanation” and “this is what you have” can determine whether they seek professional care.
When reassurance becomes a risk
Doctor Mboglen Diane recognises that AI can help people learn about diseases, understand symptoms and identify questions to raise with health professionals. But she draws a clear line between information and confirmation.
“Just as AI gives you a medical guess, a doctor runs tests. A doctor confirms. AI does not confirm,” she said.
A chatbot’s response is based on the information available to it. It may lack vital signs, examination findings, laboratory results or a patient’s full medical history. Even when it asks follow-up questions, it may not gather enough information to distinguish a minor problem from a serious condition.
Doctor Diane warns that an inaccurate or overly reassuring response could delay a visit to a health facility.
“It could cost you your life,” she said, referring to the potential consequences of errors in healthcare.
The danger arises when users interpret chatbot responses as permission to ignore symptoms, postpone treatment or take medication without professional guidance.
AI can suggest possible explanations, but it cannot independently establish that a condition is harmless. Nor does a symptom disappearing later prove that the initial assessment was correct.
The young woman’s experience ended reassuringly, but it does not establish that the same approach would be safe for someone else with similar symptoms. Another person may have a different condition or symptoms requiring urgent attention.
The gap AI is stepping into
The appeal of AI-assisted health information cannot be separated from the circumstances in which people seek care. Someone may want immediate answers but face financial constraints, difficulty reaching a health facility or discomfort discussing a private concern.
Midwife Che Mariam Mbong says people increasingly turn to AI because they already use it for everyday questions and expect immediate responses. Access to healthcare professionals is not always straightforward, and arranging a consultation can take time.
“You might be going through something that you want an immediate response to,” she explained.
Mariam says chatbot responses depend on the information users provide and cannot fully account for individual medical circumstances. Advice that appears reasonable in general may not be suitable for a particular patient.
She believes the answer lies in public education rather than pretending the technology will disappear.
“AI is not going to stop today. So what we have to do is change the narrative around the use of AI,” she said.
She advocates helping people understand what these tools can and cannot do, while encouraging developers and health organisations to strengthen safeguards against unsafe medical guidance.
Telling people not to use AI may have limited effect if they continue to need information that is private, affordable and readily available. The challenge is to help them use it without allowing it to replace appropriate medical assessment.
What research in Cameroon tells us
A community-based study published in the Journal of Tropical Medicine in 2025 examined knowledge and preparedness for adopting AI and Digital One Health technologies for home healthcare. Researchers surveyed 1,625 adults across 33 communities in the Buea, Limbe, Bonassama and New-Bell health districts between April and May 2024.
The study found that 518 participants, or 31.9 per cent, had heard of AI. Only 166, representing 10.2 per cent of the sample, demonstrated good knowledge of AI and Digital One Health technologies, according to the researchers’ criteria. Meanwhile, 941 participants, or 57.9 per cent, were classified as highly prepared to adopt the technologies studied.
The findings suggest a possible gap between willingness to adopt digital health technologies and the knowledge needed to use them effectively.
However, the study did not measure how many young Cameroonians use AI to assess symptoms, whether its responses are accurate or how often AI influences decisions to seek medical care. It should not be interpreted as evidence that chatbot diagnosis or unsafe self-medication is widespread.
Instead, the research raises an important question: as interest in AI-assisted healthcare grows, are people receiving enough guidance to understand its limitations?
WHO raised similar concerns in its 2024 guidance on the ethics and governance of large multi-modal models, warning that generative AI can produce false, inaccurate, biased or incomplete information that may harm people relying on it to make health decisions.
The WHO also recognised potential benefits for healthcare while emphasising the need for appropriate oversight and responsible use.
A tool that sounds like an expert
AI expert Vianney Forewah says one of the central concerns is the gap between what a chatbot can say and what it can establish.
Although AI platforms have introduced safeguards intended to limit inappropriate medical advice, Forewah says these vary. Some systems may still respond to questions about illnesses or medication in ways users could interpret as professional guidance.
AI works with the information available to it, which may omit details that matter to a trained practitioner. It cannot independently perform the clinical checks required to confirm a condition or determine the most appropriate treatment for an individual.
Forewah is particularly concerned that AI could compound the risks of self-medication if people use chatbot responses to decide which medicines to take or whether to visit a hospital.
“AI can give you a lead or an idea about your health, but you still need to follow the lead and go to a hospital, get properly checked and then get recommendations from health practitioners,” he said.
AI can help users understand unfamiliar medical terms, explore general information and prepare questions for consultations. The danger comes when it is treated as though it has examined the patient, verified the facts and accepted responsibility for its recommendations.
Chifen Sama Nduma, a nurse and founder of LifeDrop, a WhatsApp-based chatbot designed to facilitate blood donation, says trust is another challenge for locally developed health technologies.
Drawing on his experience, he says some users are reluctant to disclose sensitive information to platforms associated with local organisations because they worry about how their information might be accessed or used.
“People tend to lie,” he said, explaining that such concerns can affect how openly users share information. In his experience, some people are more willing to disclose confidential health concerns to global platforms such as ChatGPT than to chatbots operated by organisations in their communities.
This raises a difficult question for local developers: how can a system provide relevant support if users do not feel comfortable sharing the information it needs?
Building trust requires explaining how personal information is handled and ensuring users feel safe using these services.
Building AI for local realities
Sama argues that AI health systems must also account for the populations they serve rather than assume that general information will work equally well in every setting.
“ChatGPT is not your doctor. Go to the hospital,” he advised.
He believes AI should be developed and used responsibly, with qualified people reviewing its outputs and intervening when necessary.
A health professional can ask additional questions, examine a patient, order tests and interpret findings in context. A chatbot cannot independently carry out that clinical process.
Sama’s experience with LifeDrop illustrates a more specific way chatbot technology can support healthcare: helping people connect with a potentially life-saving service.
He says instead of circulating requests through multiple groups and contacting friends, people seeking blood can use LifeDrop to identify eligible donors and connect with them.
According to Sama, the system has reduced the time needed to find a donor from more than three hours to as little as 15 minutes in some cases, although he also cited a turnaround of about an hour in others. He reports that LifeDrop facilitated more than 27 blood donations since its launch in August 2026.
The example points to a practical role for AI: supporting specific healthcare needs without pretending to replace medical professionals.
Knowing when an answer is not enough
For the young woman who asked ChatGPT about her discharge, the experience brought relief. For Aurelia, AI provides a starting point for exploring health concerns. Health professionals, meanwhile, warn that the same convenience can become dangerous if it encourages people to delay care or act on advice that has not been properly assessed.
AI may help someone understand a medical term, recognise that a symptom deserves attention or prepare for a consultation. It may also provide incomplete information, overlook important context or offer reassurance without a sound clinical basis.
The safer approach is to treat AI as a starting point, not a final diagnosis. People should not rely on chatbot responses alone to decide that symptoms are harmless, start or stop medication, or replace recommended medical care.
As AI becomes more familiar in everyday life, healthcare professionals, technology developers and public educators face the challenge of ensuring that access to information does not come at the expense of safe care.
For a young person staring at a screen and wondering what is happening inside their body, an answer may arrive in seconds. The harder question is knowing when that answer is not enough.