Using AI in medicine is already part of everyday life. Neural networks help doctors sort electronic patient records, analyze scans, and remind people about a scheduled check-up.
A fitness tracker will always remind you that movement is life, a health app will help you track your own sleep, and a chatbot will explain confusing terminology from tests and reports.
However, an AI doctor is not a full replacement for a professional physician. Its conclusions should never be treated as absolute truth. AI is an informational assistant that can suggest when a consultation with a doctor is truly necessary.
Let's look at how neural networks work in medicine in practice, using tools from the GPTunneL hub as an example.
Case 1. GPT-5 – "Cold or flu?"
Prompt:
"Temperature 38.3 °C, cough present, no runny nose, headache and chills. What signs distinguish a cold from the flu?"
Result
Based on the description provided, GPT determined that the symptoms resembled the flu more than a common cold.
The neural network explained how three illnesses — flu, common cold, and COVID-19 — can present, and by which signs to tell them apart. The AI also gave neutral, safe recommendations: get plenty of sleep, drink warm fluids, limit contact with others, especially people in risk groups (the elderly, pregnant people, and those with serious chronic conditions). In addition, the AI listed the warning signs that call for seeing a doctor right away.
You can review the GPT output here.
Case 2. Perplexity AI – "How to read pulse oximeter readings"
Prompt:
"My SpO₂ is 96%, pulse 78. I measured it in the morning, sitting still, no nail polish, no movement. What do these numbers mean, and how do I know if they're within normal range? When does a pulse oximeter reading mean it's time to see a doctor, and how do I take readings correctly so they're more accurate?"
Result
Perplexity Sonar Pro explained in detail that an SpO₂ of 96% and a pulse of 78 bpm are within normal range (95–100% and 60–90 bpm, respectively). The neural network flagged the thresholds that warrant concern (≤94% – see a doctor, ≤90% – emergency care). The AI also pointed out what can distort the readings:
- cold fingers;
- nail polish;
- bright light;
- movement.
The user received a step-by-step guide on how to take measurements, along with a reminder that a pulse oximeter cannot replace a doctor or provide a diagnosis.
Perplexity backed up its answer with plenty of links supporting the information it provided.
You can check out this example of AI in medicine via this link.
Case 3. Gemini 2.5 Pro – "Decoding a complete blood count"
Prompt:
"I got my complete blood count results, but without any explanation. Hemoglobin – 138, hematocrit – 41, white blood cells – 6.2, platelets – 250, MCV – 87, MCH – 30. Explain in simple words what each value means, what the normal ranges are, and how to tell if everything is fine. Don't give a diagnosis — just help me understand so I know what to ask my doctor."
Result
Gemini 2.5 Pro explained, in language accessible to a non-specialist, the function of each blood component, outlined the normal ranges, and drew a conclusion about the patient's blood status. The neural network offered simple comparisons, such as "hemoglobin is like a taxi that delivers oxygen to its destination."
Gemini also suggested a set of questions the patient could ask their doctor. The model repeatedly noted that its answer is not equivalent to a specialist's consultation and that reference ranges can vary between labs.
You can see how Gemini performed by reviewing this conversation.
Case 4. DeepSeek 3.1 – "Analyzing a week of blood pressure readings"
Prompt:
"I have a week of blood pressure data and want to know if everything is okay and whether I should change any habits. Here are my measurements:
- Oct 25 – morning 122/80, evening 125/82
- Oct 26 – morning 118/78, evening 120/80
- Oct 27 – morning 126/82, evening 130/84
- Oct 28 – morning 120/79, evening 122/80
- Oct 29 – morning 119/77, evening 121/79
- Oct 30 – morning 124/81, evening 126/82
- Oct 31 – morning 123/80, evening 124/81.
Calculate the averages, compare them to the AHA classification, and explain which category they fall into. Show whether there are any trends (morning/evening, stability), and suggest general lifestyle tips — diet, sleep, movement, stress. Don't give medical prescriptions, just help me understand what could be improved to keep my blood pressure normal."
Result
DeepSeek 3.1 accurately calculated the averages: 123/80 mmHg (morning 121/79, evening 124/81). Guided by AHA data, the neural network classified the readings as elevated, with a slight risk of developing hypertension. The AI also identified a trend — a mild rise in blood pressure toward the evening.
DeepSeek also gave recommendations for reducing risk: cut down on salt intake, add outdoor walks, improve sleep, and try to manage stress better.
You can gauge the accuracy of AI in the medical field by checking out this chat.
Case 5. LLaMA 4 Scout – "Sleep analysis and rest-hygiene tips"
Prompt:
"I keep a sleep diary and want to know whether I'm getting enough sleep and what could be improved. Here's my data for the week:
- Oct 25 – bedtime 00:30, wake-up 07:30, awakenings 1, coffee after 3 PM – yes, screen before bed – yes
- Oct 26 – bedtime 00:45, wake-up 07:45, awakenings 2, coffee after 3 PM – no, screen – yes
- Oct 27 – bedtime 23:50, wake-up 07:10, awakenings 0, coffee after 3 PM – yes, screen – no
- Oct 28 – bedtime 00:15, wake-up 07:20, awakenings 1, coffee after 3 PM – no, screen – yes
- Oct 29 – bedtime 00:40, wake-up 07:30, awakenings 1, coffee after 3 PM – yes, screen – yes
- Oct 30 – bedtime 00:00, wake-up 07:00, awakenings 0, coffee after 3 PM – no, screen – no
- Oct 31 – bedtime 00:20, wake-up 07:15, awakenings 1, coffee after 3 PM – no, screen – yes
Calculate the average sleep duration, assess whether it's enough for an adult, and explain how evening coffee and screen time before bed affect sleep. Give 3–5 tips for improving sleep hygiene based on CDC and Harvard Health guidance. Speak in a simple, friendly tone, like a sleep consultant, and add a short summary: what I'm doing right and what's worth adjusting."
Result
LLaMA 4 Scout calculated an average sleep duration of about 7 hours. The neural network explained that for adults, this is the lower end of the normal range (7–9 hours is optimal). The AI also drew the user's attention to the fact that blue light from device screens and caffeine negatively affect melatonin production and sleep quality.
To make sleep deeper and longer, LLaMA 4 Scout advised the user to avoid coffee 4–6 hours before bedtime and to put the smartphone away about an hour before sleep.
You can read the neural network's full response by following this link.
Conclusion
Today, neural networks in medicine are moving from novelty to everyday tool. They can explain medical terms in plain language, decode test results and scans, offer neutral recommendations and instructions, and flag when it's time to see a doctor.
When using the capabilities of smart models, the key is to avoid self-diagnosis and self-treatment with AI. A neural network relies on general, industry-wide knowledge, while every real patient is unique and requires an individual approach that only a specialist can provide.