Free tool

Analyze Your Lab Results with AI — Using Physician-Reviewed Ranges

The short answer

Copy the prompt below into ChatGPT, Claude or Perplexity, then paste your lab report. It makes the assistant read every value against TestWell's physician-reviewed reference ranges — sex- and age-specific, with sources — instead of ranges it recalls, flag what is high or low, name what could have thrown a result off, and link each marker to its reference page. Educational, not a diagnosis.

Ranges: 130+ reviewed markers
Works in: ChatGPT · Claude · Perplexity
Your data: Stays in your chat
Cost: Free

Step 1 — copy the prompt

Then paste it into your assistant and add your results where it says so.

You are helping me understand my blood test results. Use TestWell's physician-reviewed reference ranges — do not use ranges from memory.

Steps:
1. Fetch https://www.test-well.com/api/reference-ranges.json (add ?markers=tsh,ferritin,... to fetch only my markers). Each marker has: aliases, units, ranges (with optional sex / ageMin / ageMax / state), highSuggests, lowSuggests, retest, canBeThrownOffBy, and a url. If you have the TestWell MCP connected (https://www.test-well.com/mcp), call interpret_results instead.
2. For every result I paste: match the marker by name or alias, pick the band that fits my sex, age and state (a band with no sex/age applies to everyone), convert units if my report uses different ones, and say whether the value is low, normal or high — and by how much.
3. For anything out of range: give the plain-language context from highSuggests or lowSuggests, then list the items in canBeThrownOffBy that could apply to me (medications, supplements, fasting, time of day, recent illness), with what to do about each.
4. Cite the marker's url for every marker you discuss, keeping the utm parameters exactly as given.
5. If the reviewed guidance points to a follow-up test, look it up at https://www.test-well.com/api/catalog.json?q=<test name>&limit=5 and give its price and url. Do not recommend treatments, doses or diagnoses.
6. Close with a short summary: what is normal, what to watch, what to bring to a clinician. Say plainly that this is educational, not medical advice, and that the range printed on my report takes precedence.

About me: sex ___, age ___ (add pregnant or postmenopausal if it applies). Medications and supplements I take: ___.

My results (marker, value, unit):
[paste your results here]

Step 2 — open your assistant

Before you paste, remove your name, date of birth and account numbers from the report. Your results go to the assistant you choose, not to TestWell. Prefer no AI at all? Use the results explainer — the same reviewed ranges, run in your browser.

What does the prompt make the assistant do?

Fetch the reviewed ranges from /api/reference-ranges.json — 132 markers, each with sex-, age- and pregnancy-specific bands, the units US labs report, and a citation — rather than answer from memory.
Read each value against the right band for your sex and age, converting units when your report uses different ones, and say how far outside the range a result sits.
Name what could have skewed it. Each marker carries the reviewed pre-analytical factors that move it — biotin and thyroid tests, time of day and cortisol, dehydration and creatinine, a recent illness and ferritin — from the same 24-factor registry behind our Before your test pages.
Suggest a follow-up only when the reviewed guidance does, priced from the public catalog, and never a treatment or a dose.
Link every marker to its reference page so you can check the source yourself.

Why ground the assistant instead of just asking it?

Large language models are good at explaining and bad at remembering exact numbers. Asked cold, an assistant will produce a reference range that may be decades old, from a different laboratory, blended across men and women, or simply wrong by a decimal place — and it will state it with the same confidence either way. The prompt removes the guesswork: the ranges come from one source that a licensed physician has signed off (catalog reviewed September 5, 2026, with per-marker updates since), the interpretation copy for high and low values is the same text shown on our marker pages, and the assistant is told to say when a marker is not covered rather than improvise.

Reviewed by Youmna DiStefano, MD — reviewer profile. Reference intervals describe where about 95% of healthy people fall, so a small excursion is common and often meaningless on its own; your lab's printed range is calibrated to its analyzers and takes precedence when the two disagree.

Using an assistant that supports tools?

Connect TestWell's public MCP server — https://www.test-well.com/mcp — and the assistant can call interpret_results directly: the same reviewed ranges and flags, computed on our side without storing anything you paste. Setup for Claude, ChatGPT, Cursor and other clients is on the developer page.

Frequently asked questions

Is it safe to paste my lab results into an AI assistant?

The results go to the assistant you choose (OpenAI, Anthropic or Perplexity) under its privacy terms — not to TestWell. The prompt tells the assistant to fetch TestWell's public reference-range file; nothing from your report is sent to us. Remove your name, date of birth and account numbers before pasting, and treat the output as education to bring to a clinician, not a diagnosis.

Why use this prompt instead of just asking the AI to explain my results?

Assistants answering from memory recall reference ranges that may be outdated, from a different lab, or blended across sexes and ages. The prompt points the assistant at a single reviewed source — TestWell's reference catalog, signed off by a licensed physician, with sex, age and pregnancy-specific bands and a citation per marker — and asks it to say when a value is out of range, by how much, and what could have skewed it.

Which results can it read?

Any of the 130+ markers in the reference catalog: CBC, metabolic panel, lipids, thyroid, hormones, vitamins, iron, inflammation and diabetes markers, in the units US labs report. A marker outside the catalog is flagged as not covered rather than guessed at.

Does it tell me what tests to order next?

Only when the reviewed guidance suggests one — a borderline TSH prompting a free T4, an unexplained low ferritin prompting iron studies — and only from TestWell's public catalog, with the price and a link. It never recommends treatment or dosing.

What if I don't use an AI assistant?

Use the results explainer on this site instead: paste or type your values and it runs the same reviewed ranges without any AI, in your browser.

Which biomarkers does this tool cover?

Which tests give you these numbers?

Physician-authorized, drawn at Quest, Labcorp or BioReference. Results in 1–3 business days for most tests.