AI chatbot for health questions: how much can you trust the answer?
ca. 18 Minuten
The essentials
An AI chatbot offers guidance; it doesn’t diagnose. It can explain what a symptom might indicate, which questions are useful to ask at a doctor’s office, and how a lab result is obtained. It can’t know what’s going on with you: it doesn’t have your medical history, an examination, or any measured values. The line is crossed when an explanation becomes a diagnosis.
This isn’t a warning about the technology, but a description of how it’s built. A language model generates text that fits your question. Whether that text is correct is a separate question, and it isn’t answered during generation. If you know this, you can use a chat window sensibly. If you don’t, you may mistake a fluent answer for a verified one.
Chapters 1 and 2 explain what actually happens when the model answers. Chapter 4 draws the line between guidance and self-diagnosis. Chapter 6 reveals where an AI provides advice in this shop and what it can’t do. Chapter 7 describes situations where a chat window is no longer the right place to turn.
What to expect in this article
1. What a chatbot does when you describe a symptom
2. Why a wrong answer sounds just as confident as a right one
3. Three things people say about AI answers that aren’t true
4. Guidance or self-diagnosis: the distinction in detail
5. What can’t be replaced by a language model is what gets measured
6. The AI in our shop: what Cloé can and can’t do
7. When the question belongs in a doctor’s office, not a chat window
8. How to ask without going down the wrong track
9. Who benefits from the chat window and who doesn’t
10. What you can do differently next time
Frequently asked questions
Sources
What a chatbot does when you describe a symptom
You type three lines about fatigue, upper abdominal pressure, and poor sleep. Two seconds later, there’s a neatly structured text with possible causes, recommendations, and a friendly closing sentence. What happened in those two seconds had nothing to do with examining you.
A large language model calculates which word is most likely to follow the previous one. It has read a great deal of text, including medical literature, forum posts, advertising copy, and advice pages. From this, it forms an answer that fits your question. This process does not include a verification step that checks the statement against a source.
An image for this: The model works like an exceptionally well-read night porter. He has read every magazine in the building, knows the familiar phrases on every subject, and phrases them calmly and clearly. He has never examined you, and he will never say that he doesn’t know you.
Key takeaway
A language model assesses which words fit together. It doesn’t check whether a statement is true.
How widespread chatbot use has become
You’re not alone in having this question. According to a representative survey of 1,145 people aged 16 and over, 45 percent of people in Germany use AI chatbots to ask about symptoms and health; of those who do, 55 percent trust the answers (Bitkom e. V., 2025). That means the chat window is the first place almost one in two people describe a health concern.
That’s understandable. A chatbot is available at three in the morning, it doesn’t interrupt, and it requires no preparation. Those three qualities make it a good starting point and a bad endpoint.
Why a wrong answer sounds just as confident as a right one
With a person, you can recognize uncertainty from hesitation. A language model doesn’t hesitate. It presents a fabricated study citation in the same calm tone as a correct one. The technical term for this is hallucination, and it describes not an exceptional case, but a feature of how the system is designed.
Verified source
“If the AI has no data that matches the user’s question, it may make up the answer. This is also called a hallucination. The answer may sound plausible, but be entirely fabricated. The same applies to citations.”
Health Knowledge Foundation
Searching for health information with AI chatbots, as of 2025
The sentence about citations is the most unsettling. If you’re skeptical, you can check a fabricated claim. A fabricated source behind the claim takes away that very skepticism because it looks like evidence. If you don’t open the source, you think the answer is doubly verified.
The second error happens when reading
In 2024, the World Health Organization published guidance on large multimodal models in healthcare. It warns against statements that are “false, inaccurate, biased, or incomplete” and may harm people, and it identifies a second mechanism: “automation bias,” the tendency of healthcare professionals and patients to overlook errors in machine-generated answers (World Health Organization, 2024).
That means the text doesn’t seem convincing because it’s good, but because it comes from a machine and shows no obvious doubts. If a person told you this, you’d ask where they got it from. When the answer has no face, people don’t ask.
Three statements about AI answers that aren’t true as stated
The following three assumptions regularly come up in conversations about chatbots. Each is understandable, and each leads you astray in a different way.
Common assumption
If the chatbot cites sources, they’ve been checked.
What actually applies
Sources can be just as fabricated as the sentence before them. The Stiftung Gesundheitswissen explicitly states this in 2025. A source can only be verified once you open it and find the quoted sentence there.
Common assumption
A detailed answer is a reliable answer.
What actually applies
Length comes from the amount of matching text patterns, not certainty. A model often has more to say about a rare set of symptoms than a common one because rare conditions are described in more detail in medical literature. So length says nothing about accuracy.
Common assumption
If I type in my results, the AI interprets them correctly.
What actually applies
A model knows reference ranges as text, not the ones from your lab. Reference ranges depend on the lab, method, and age. A value identified as abnormal in an answer may fall within the range stated on your test results.
Guidance or self-diagnosis: the line in detail
The line isn’t a feeling; you can tell from the question you ask and what you do with the answer. The same conversation can start on one side of the line and end on the other without the tone changing.
| Criterion | Guidance – holds up | Self-diagnosis – doesn’t hold up |
|---|---|---|
| Your question | What could this symptom indicate? What tests are typically done for it? | What do I have? Is it dangerous? Do I need a doctor or not? |
| What you get back | Terms, connections, possible tests, vocabulary for your next conversation | An assessment that looks like a diagnosis but isn’t based on any examination |
| Data basis | General knowledge about medical conditions, independent of you | Your typed description, without an examination, medical history, or test results |
| What you do afterward | You write down three questions for the appointment and go | You postpone the appointment, stop taking a medication, or start one on your own |
| How you can tell | You have more questions than before, but better ones | You feel relieved or scared, but don’t have a single new question |
| The honest test | You could show the answer to a doctor without having to justify yourself | You’d rather not mention in the conversation where you got it |
The last line is the most useful. If you wouldn’t take an answer openly into the doctor’s office, you’ve already treated it as a substitute, not as preparation.
What’s measured can’t be replaced by a language model
There’s a difference between a statement about ferritin and your ferritin level. The first appears in thousands of texts a model can draw on. The second exists only after someone has taken a blood sample and a lab has analyzed it.
A language model knows sentences about your body. It doesn’t know a single value from it.
That’s why a suspected deficiency should be investigated with a test, not a conversation with a machine. We’ve compiled the symptoms that may warrant testing in our article on symptoms of nutrient deficiencies. Our article what a complete blood count examines explains what a blood test covers.
The reverse is just as true, and it’s the more uncomfortable part. Even a measured value doesn’t answer the question you came with. It’s a data point, not a verdict. The article which diseases can be identified in a complete blood count explains which conditions a blood test can and can’t indicate; the same logic applies to the gut, as the article testing the microbiome shows.
An example from our own product range to give you a sense of scale: the VitalCheck | The Complete Nutrient & Mineral Test from mybody®x (MYBODY Lab GmbH) measures 18 biomarkers from capillary blood, or just a few drops from your fingertip, and costs €169.00 (as of 03.09.2026; subject to change). It provides figures for vitamins, minerals and iron status. It won’t tell you what’s causing your fatigue, and it doesn’t replace a medical examination. A number is a starting point, not a conclusion.
The difference between a chatbot and a lab is not one of reliability, but of responsibility. One generates text, the other generates numbers. In both cases, interpretation remains with a trained human.
The shop’s own AI: what Cloé can and can’t do
An article about the limits of AI that hides its own AI would be worthless. So, in order: this shop uses an AI consultation tool. It’s called Cloé, appears as a question bar among the products, and answers questions about the health tests.
Key takeaway
Cloé advises on tests and products. She does not diagnose or read test results.
Since 2024, the European AI Act has required people to be informed when they are talking to a machine. Article 50(1) puts it this way: providers shall ensure that AI systems intended to interact directly with natural persons “are designed and developed in such a way that the natural persons concerned are informed that they are interacting with an AI system” (Regulation (EU) 2024/1689, Article 50(1)).
Cloé is identified as AI in the consultation; this identification cannot be turned off. She works from a knowledge base covering her own product range, not the open internet. Her recommendations are tests in the areas of DNA, blood, and gut health.
What Cloé explicitly does not do
Four things the consultation can’t do
No diagnosis
She provides context and narrows things down. She won’t tell you what you have or rule anything out.
No reading test results
She does not accept lab results or interpret values. A test result belongs with the doctor who ordered it.
No view beyond her own fence
Her knowledge base ends with her own product range. She provides sales advice with mandatory professional disclaimers, not independent information about the market.
No children, no plans to have children
For these two topics, she refers you to a doctor’s office and deliberately does not recommend a replacement product.
So the same applies to Cloé as to every other language model discussed in the previous chapters. She’s faster than an email to customer service and worse than a conversation with a professional. You can’t have both. Use her for the first, and you save time; use her for the second, and you lose something more important.
When the question belongs in a doctor’s office, not a chat window
There are symptoms for which every minute of research is a minute too long. The following list is no substitute for an assessment; it is a stop condition: if any line applies, end your search in the chat window and go to a doctor’s office or an emergency room.
Where self-research ends
Sudden, severe, unlike usual
Pain that comes out of nowhere and feels unfamiliar. Especially in your chest, head, or abdomen.
Blood where it shouldn’t be
In your stool, urine, vomit, or when you cough. Even just once, and even if it stops afterward.
Shortness of breath or pressure in your chest
Don’t research or wait it out. Call 112.
Unintentional weight loss or persistent fever
If you’re losing weight without trying or have a fever that lasts for days, get it checked out rather than trying to interpret it yourself.
Children, pregnancy, ongoing treatment
In these three situations, the gap between general information and your situation is widest. Never stop taking medication based on a chat response.
Nowhere in this article is the doctor the chatbot’s adversary. The doctor is the next point of contact, and a well-prepared conversation is worth more than a perfectly worded answer at midnight.
How to ask without going down the wrong path
The following four steps don’t change the technology, but your role in using it. They’re meant to help you prepare for a doctor’s appointment and are explicitly not a way to diagnose yourself.
Ask about terms, not diagnoses
Instead of “What’s wrong with me?”, try “What technical terms describe this symptom, and what is usually examined in such cases?” That way, you get vocabulary rather than a diagnosis. Source: Stiftung Gesundheitswissen, as of 2025.
Ask about the uncertainty too
Add to every question: what’s uncertain about the answer, and what argues against it? Models rarely provide this part on their own because it doesn’t look like an answer. Source: World Health Organization, 2024.
Open every source yourself
Open the cited source and look for the quoted sentence there. If you can’t find it, the claim doesn’t hold up. Fabricated citations are a documented problem, not a one-off. Source: Stiftung Gesundheitswissen, as of 2025.
Turn the answer into three questions
Write down three questions from the chat history to take to your appointment. That’s the real benefit: a better first minute in the doctor’s office. Source: mybody®x editorial team, as of 03.09.2026.
At a glance
Ask about terms, ask it to explain the uncertainty, check every source, and take three questions to your appointment.
Who benefits from the chat window and who doesn’t
The question isn’t whether a chatbot is good or bad. The question is what situation you’re in right now.
It makes sense for you if ...
... you have a test result in front of you and don’t know the technical terms in it. A model explains vocabulary well because it doesn’t need to know anything about you.
… you have an appointment and want to prepare. Sorting out questions is a language task, and language tasks are what this technology is good at.
… you want to know which tests are commonly done for a set of symptoms, so you won’t be at a loss during the conversation.
Probably not, if …
… you’re really looking for reassurance. Then you read the answer the way you need to, and your appointment gets pushed back by weeks.
… you’re making a decision about medication. Stopping, changing the dose, or starting something new should be handled by a doctor.
… one of the lines in the chapter on warning signs applies. In that case, the answer on the screen is the wrong thing to focus on.
What you can do differently next time
The common advice is to avoid chatbots for health questions. It ignores reality: 45 percent of people in Germany use them for this anyway (Bitkom e. V., 2025), and a ban won’t change that. The more useful question is what task you give the chat window.
There’s an observation from running our own advice service that helps explain this. An AI system doesn’t stay quiet when it lacks information. It keeps answering, but without a basis, and no one can tell from the answer. At Cloé, this once meant she gave detailed advice but stopped recommending any products because her product list no longer matched the shop. The text was flawless; the foundation was gone.
Applied to your health question, that means the risk isn’t the wrong answer, but the missing gap in the text. A person says, “I don’t know.” A model fills in the blank.
So next time, take two minutes to pull three questions for your appointment out of the chat. That’s the whole trick. The consultation in the shop is free, and you won’t get a diagnosis there.
Frequently asked questions
Can an AI chatbot diagnose me?
No. A chatbot generates text based on probabilities; it doesn’t examine you or know any of your test results. In 2025, the Stiftung Gesundheitswissen noted that while you’ll always get an answer to a question about an illness, it isn’t necessarily correct, and that a chatbot therefore can’t replace a visit to the doctor. What it can do is help you prepare: explain terms, name possible tests, and help you come up with the right questions to ask in the doctor’s office.
How can I tell if an AI answer is made up?
You can’t tell from the text itself, and that’s the heart of the problem. A fabricated answer can be worded just as fluently as a correct one. The only reliable way is to check the source: open the cited source and look for the quoted sentence there. If you can’t find it, or the page doesn’t exist at all, the claim doesn’t hold up. In 2025, the Stiftung Gesundheitswissen explicitly noted that sources can also be made up.
Can I type my lab results into a chatbot?
Technically, you can, but it’s only advisable to a limited extent. A model knows reference ranges as general text, not the ranges used by your lab. Reference ranges depend on the lab, method and age; the information on your lab report is always authoritative. There’s also the issue of data protection: health data is especially protected, and anything you enter into an external system is beyond your control. The practice that ordered the test is responsible for interpreting your results.
Does it have to be clear that I’m talking to an AI?
Yes. Article 50(1) of the European AI Act requires AI systems designed to interact directly with natural persons to be designed so that those persons are informed that they are interacting with an AI system. The exception applies only when this is obvious from the circumstances anyway.
Cloé, the advisor in this shop, is labeled as AI; the label cannot be turned off.
What can the AI advisor Cloé do, and what can’t it do?
Cloé answers questions about health tests in the areas of DNA, blood and gut health, and recommends suitable products based on them. It works with a stored knowledge base covering its own product range, not the open internet. It does not diagnose, read or interpret lab results, or make claims about providers outside its own company. For topics related to children and trying to conceive, it refers users to a medical practice and deliberately does not recommend a substitute product.
Next step
Turning a guess into a measurement
If, after the chat, you want to know where you actually stand, a measurement is more useful than the next question. You can then discuss what follows with a person.
View VitalCheck Complete All articles about blood testsRead more
You might also be interested in
What preventive care can do when it’s based on more than guesswork.
An example of how far individual values can take you and where their limits lie.
Sources
- Stiftung Gesundheitswissen: Searching for health information with AI chatbots (as of 2025) – stiftung-gesundheitswissen.de
- World Health Organization: WHO releases AI ethics and governance guidance for large multi-modal models (as of 2024) – who.int
- Bitkom e. V.: Dr. AI – When the chatbot becomes a medical adviser (as of 2025) – bitkom.org
- European Union: Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence, Article 50 (as of 2024) – eur-lex.europa.eu
The statements about hallucinations, fabricated sources, and chatbots not replacing a doctor’s visit are based on [1]. The warning about false, inaccurate, distorted, or incomplete statements, as well as the term automation bias, comes from [2]. Usage and trust figures for Germany come from the representative survey of 1,145 people aged 16 and over in [3]. The wording on labeling requirements is from [4]. The product name, price, and number of markers were checked on mybody®x product pages on 03.09.2026.
mybody®x Editorial & Expert Team
Digital health information Laboratory diagnostics AI in the shop
This article was written by the mybody®x editorial team, with the specialist team reviewing the laboratory diagnostic statements. The contributors are listed on the authors page.
Published on 03.09.2026 · Last updated on 03.09.2026
The content is for general information and is not a substitute for medical advice, diagnosis, or treatment. Reference ranges depend on the laboratory, method, and age – the information on your test result is always authoritative. Our products are not intended to diagnose, treat, cure, or prevent diseases.
