In short
- A good ChatGPT blood test results prompt gives your age, sex, each value with its unit and your lab's own range, plus medicines, fasting and pregnancy or cycle.
- Tell the model not to diagnose, to say how certain it is, to name a source for each range and to list questions for your doctor.
- Never paste your name, birth date, ID or insurance number, address, or doctor and lab ID numbers. The numbers explain the result; your identity does not.
- Some values need a partner. Ferritin rises with inflammation, so give ferritin together with CRP.
- A prompt lowers errors but does not remove them. Chatbots can still guess ranges, slip on units and answer the same question differently.
A bare "explain my blood test" gets a generic answer. Below you get a prompt template to copy, a worked example with two values that must be read together, and an honest list of what no prompt can fix. For the wider picture, start with what an AI blood test analyzer can and cannot do.
Can ChatGPT interpret blood test results?
Yes, ChatGPT can interpret typed blood test results in plain words, but the quality depends on what you give it. Without your age, sex, units and your lab's range, it fills the gaps from memory, and that is where most mistakes start.
A 2024 study in the Journal of Medical Internet Research rated GPT-4's answers to lab questions best among the models tested. The authors also found that some answers were inaccurate and not individualised. A 2026 study in npj Digital Medicine shows why context matters: accuracy rose from 0.80 to 0.92 when models were given reference ranges.
What should a ChatGPT blood test analysis prompt include?
A good prompt includes eight facts about you and your sample. Each one can change how a value is read, and a missing one forces the chatbot to guess.
| What to include | Why it matters | Example |
|---|---|---|
| Age | Some ranges change with age, so a value can be fine at one age and not at another. | 46 years |
| Sex | Some cutoffs differ by sex. WHO sets anemia cutoffs for hemoglobin at 120 g/L for women and 130 g/L for men. | Hemoglobin 125 g/L, female |
| Each value with its unit | The same result looks different in different units, and a missing unit forces a guess. | Hemoglobin 125 g/L (12.5 g/dL) |
| Your lab's reference range | Labs use different methods, so their ranges differ. Compare with your own lab's range. | TSH 4.6 mIU/L, lab range 0.4–4.0 |
| Medicines and supplements | Medicines and supplements can affect some results. | Iron tablets for 8 weeks |
| Fasting and time of the draw | Food, drinks and hard exercise can affect some results. | Fasting 10 hours, blood taken at 8:00 |
| Pregnancy or cycle phase | Menstrual periods can affect results, and pregnancy changes how some values are read. | Not pregnant, cycle day 3 |
| Recent illness or injury | Inflammation markers such as CRP rise for a while during an infection. | Cold with fever last week |
Source: hemoglobin cutoffs from WHO 2024; TSH band 0.4–4.0 mIU/L from the European Thyroid Association 2013 guideline. Factors that affect results are from MedlinePlus (NIH), How to Understand Your Lab Results. Reference ranges differ between laboratories.
A copy-ready ChatGPT prompt for lab results
Copy the template below, fill in the square brackets, and delete any line that does not apply to you.
I am sharing my own blood test results. Please explain them in plain language. About me: age [years], sex [female or male]. Pregnant: [no, or yes and which week]. Cycle day: [number, or not relevant]. Medicines and supplements: [names and doses, or none]. Blood taken: [date and time]. Fasting: [yes and how many hours, or no]. Recent illness, injury or hard exercise: [describe, or none]. Results, each with its unit and my lab's reference range as printed: [Marker]: [value] [unit] (lab range [low]–[high]). [Marker]: [value] [unit] (lab range [low]–[high]). Rules for your answer: 1. Use my lab's ranges. If a range is missing, say so and do not invent one. 2. Do not diagnose. Explain what each value can mean and what else could explain it. 3. Point out values that should be read together, such as ferritin with CRP. 4. Say how certain you are. Say clearly when you do not know. 5. Name the source for each range or claim, such as a guideline and its year. 6. If a unit looks unusual, ask me instead of converting it silently. 7. End with a short list of questions I can ask my doctor.
Typed values are easier to check than a photo of the report. Before you send, compare every number and unit with the paper report. Then run the same prompt in a new chat. If the two answers differ in substance, take the question to your doctor.
What should you never paste into ChatGPT?
Never paste anything that identifies you, because the answer does not need it. Remove these before you type or upload.
- Your full name.
- Your date of birth. Give your age in years instead.
- Your national ID number or health insurance number.
- Your home address, phone number and email.
- Your doctor's name and ID, the lab's ID numbers, the sample number and any barcode or QR code.
If you upload a PDF or a photo instead of typing, black out these fields first. Our guide is it safe to upload blood test results to ChatGPT covers training settings, Temporary Chat and how long files are kept.
Worked example: why must ferritin be read with CRP?
Ferritin shows your iron stores, but it also rises during inflammation. A ferritin value read without CRP can therefore look fine while your stores are low.
Take a sample report: a 34-year-old woman with ferritin 45 µg/L and hs-CRP 8 mg/L. On its own, ferritin 45 µg/L sits in the band our knowledge base rates as within range. That band is 30–200 µg/L, following WHO 2020. Below 15 µg/L is low, and 15–30 µg/L is borderline.
The hs-CRP changes the picture. Our bands, from AHA/CDC 2003, rate below 1 mg/L as low, 1–3 mg/L as moderate and above 3 mg/L as raised. hs-CRP measures the same protein as a standard CRP test, only more precisely at low levels. WHO 2020 sets a higher ferritin threshold when inflammation is present. Our knowledge base follows it: with CRP above 5 mg/L, ferritin up to 70 µg/L can still point to depleted stores.
A good answer to this pair says that ferritin looks within range, but inflammation may hide low stores. It suggests repeating both tests when you are well, and asking about transferrin saturation. A weak answer says "ferritin normal, nothing to worry about". Rule 3 in the template is there to catch exactly this.
The same logic applies to TSH and hemoglobin: without your lab's range and your sex, the chatbot has to guess.
What can a prompt not fix?
A good prompt reduces errors, but it cannot make a chatbot consistent, certain about ranges or immune to unit mistakes.
Guessed ranges.
In a 2025 study in Frontiers in Artificial Intelligence, researchers sent 726,000 requests for reference intervals to ChatGPT. The lower limit it gave varied across repeats, with a median coefficient of variation of 26.5 %. About 8 % of outputs were not in the requested format.
Uneven use of age and sex.
The npj Digital Medicine study found that accuracy rose from 0.80 to 0.92 when models had reference ranges. It also found age and sex biases in how models rated blood pressure and body mass index.
Different answers to the same question.
In a 2025 study in Frontiers in Digital Health, chatbots' consistency on repeated identical clinical questions ranged from 26 % to 68 %. Those questions were about a guideline for sciatica, not lab results. Still, a chatbot writes each answer anew, so emphasis can shift between chats.
Unit slips.
If one value is in mg/dL and the range is in mmol/L, a chatbot may convert silently or mix them up. Ask it to flag units, and check any conversion yourself. For the basics, read what blood reference ranges mean and what an out-of-range result means.
Do not act on a chatbot answer alone. Never start, stop or change a medicine or dose because of a chat. If a result comes with chest pain, breathlessness, fainting, confusion, new weakness, black stools or heavy bleeding, get medical help now. In an emergency, call 112 or your local emergency number.
How Blood Lab 360 reads this
Each value is rated by a fixed rule from a knowledge base, and every range there has a published source. The rule uses your age and sex, and the app also uses the range printed on your report. The same report gets the same rating every time. If your age or sex is missing, the app does not guess a range.
The knowledge base also records which markers belong together, such as ferritin with hemoglobin and transferrin saturation. It notes when a follow-up test makes sense; for ferritin, that is after about 8–12 weeks of therapy. The AI chat answers only from your measured values and that knowledge base. Your report is parsed in your browser, and no name is needed. If you use the chat, your measured values are sent to an AI processor, with your separate consent. None of this replaces your doctor. For a walk through a whole report, see how to read a blood test report.
Frequently asked questions
Can ChatGPT interpret blood test results?
Yes, it can explain typed results in plain words, often well. It does not know your lab's range, age, sex or medicines unless you give them. Treat its answer as a draft to discuss with your doctor.
Can free ChatGPT interpret blood test results?
Yes, the free version can read typed values and explain them. The limits are the same as in paid versions: give your lab's ranges and units, remove your identity, and do not treat the answer as a diagnosis.
What is a good ChatGPT prompt for lab results?
State your age, sex, each value with unit and lab range, medicines, fasting and pregnancy or cycle. Ask the model not to diagnose, to state uncertainty, to name sources and to list questions for your doctor.
How do I write a ChatGPT prompt to analyze lab results?
Copy the template above, fill in the brackets and check every number against the paper report. Then ask the same question in a new chat and compare the two answers.
Why should ferritin be read with CRP?
Ferritin also rises during inflammation. A normal-looking ferritin next to a raised CRP can still hide low iron stores.
