Generic answers are not a flaw in the model. They are what a model produces without context. Ask whether an LDL of 130 is high and you get the population answer, the one written for nobody in particular. Give the same model your last four LDL values, the reference range your lab printed, and the trend between reports, and the answer becomes yours. The hard part is assembling that context accurately: structured values with units and ranges, aligned across labs, linked to the source documents, and kept current. That assembly is exactly what Doctano does for you.
Why every answer sounds the same
A model's medical knowledge comes from textbooks, guidelines, and studies written about populations. When you ask it a question with nothing attached, the only responsible answer is the general one, wrapped in hedges. Hematologist Dr. Adeel Khan described it plainly in TIME: without a scenario, the answer is not specific to anyone - it is the textbook answer.
People have already noticed. In a KFF survey published in March 2026, about a third of US adults said they use AI chatbots for health information or advice, and among them, 41% had uploaded personal medical information such as test results to get more personal answers. At the same time, two thirds of adults say they do not trust AI health information. Those two numbers are the same problem viewed from both sides: people are adding context because they can feel it is missing, and they still do not trust the result, because a blob of text in a chat window is a fragile way to hold a medical history.
So the question is not whether AI can be useful for your health. It is where your context should live so that the answers are about you, verifiable, and still there next year.
What health context actually is
"Context" sounds vague. It is not. For lab results, it is made of specific things:
- Values with units. 107.6 means one thing in mg/dL and another in mmol/L. A unitless number is a guess waiting to happen.
- Reference ranges from the lab that ran the test. Ranges differ by lab, method, age, and sex. One global range will misclassify your results.
- Time. A single value says little. 4.9 on its own is trivia; 4.9 after three years at 2.1 is a story. Trends are where most of the meaning lives.
- Cross-lab continuity. The same marker arrives as SGOT and AST, PCV and haematocrit, in different units at different labs. Without aligning them, every lab change resets your history.
- The source document. When a value looks wrong, the original report is the only way to check. Context you cannot verify is context you should not trust.
- Whose data it is. For a family, context is per person. Your father's HbA1c has nothing to do with yours.
- What is missing. A trustworthy AI knows what it has not seen. If it has no thyroid panel, it should say so rather than answer from imagination.
The ways people give AI their context today
Most people already try one of three things. Each works a little, and each breaks in a specific place.
Typing numbers into a chat
Fine for one number, once. But transcription errors are easy, you rarely type the units and the reference range along with the value, and the context evaporates with the conversation. Next month you start again from nothing.
Uploading raw report files
This works in the moment, which is why so many people do it. The problems show up later. Dense multi-panel tables are exactly where extraction mistakes happen, and nothing links an extracted value back to the document, so a misread becomes an answer. Reports from different labs never get aligned, so there is no real trend. And a medical document pasted into a general-purpose chatbot is handled under consumer terms, not medical privacy law. Dedicated health spaces inside chatbots are appearing for a reason, but they connect to provider records in the US and do nothing for the printed report or the PDF you already have from a lab in India.
Connecting provider records
Automatic and authoritative where it works: one organization's records, on their terms, often with a limited window of history. It cannot ingest a PDF from another lab, a photo of a printed report, or anything from a provider that does not participate.
| What matters | Typing into chat | Uploading raw files | Provider record sync | Doctano |
|---|---|---|---|---|
| History across reports | One message | Sits in one chat | Per provider | One timeline |
| Normalizes across labs | No | No | One organization | Names and units |
| Units and reference ranges | Only if you type them | Whatever is readable | As supplied | Per result |
| Value traced to its source | No | The file, if kept | Record level | Every value |
| Works with any lab's report | Partly | Yes | Participating only | Yes |
| Stays current | Retype each time | Re-upload each time | Automatic where connected | One upload |
| Knows what it has not seen | No | Not reliably | No | Marks unknown |
What changes when the context is there
Same model, different inputs, different answer. Watch the same question with and without context:
Without context
"Is an LDL of 130 high?" The honest answer is the population one: above most labs' optimal band, worth discussing, here is what LDL is. Correct, general, and identical for everyone who asks. Nothing in the answer belongs to you.
With context
"My LDL went from 96 to 107.6 across my last three reports, and my lab calls anything over 100 borderline. What changed?" Now the model can reason about your trend, your lab's own range, and what is worth bringing up at your next appointment. The answer has a direction, a timeframe, and a source you can check.
Questions that only work with context are the useful ones: what changed since my last report, which results moved outside their range, what is worth asking my doctor about. Without context they collapse back into textbook answers.
One honest caveat before the sales pitch: context makes answers relevant, not medical care. An AI with perfect context can still be wrong, which is why every value should stay linked to the document it came from, and the interpretation belongs to a qualified professional.
How Doctano builds the context
Doctano exists to do the assembly work that makes AI answers personal. The pipeline:
- You upload what you have. A PDF, a scan, or a phone photo of a printed report, up to 50 MB each, several at once, from any lab or region.
- Doctano reads it. Biomarkers, values, units, reference ranges, panels, and report dates are extracted and mapped to LOINC codes, the standard vocabulary for lab tests.
- Names and units are aligned. The same marker under different names and units at different labs becomes one series on one timeline.
- Each result is classified. Against the range printed by the lab that ran that test: healthy, borderline, critical, or unknown.
- Everything links back. Any value on the timeline opens the report it came from, so extraction can be checked instead of trusted.
- DoctanoAI answers from that context. Your latest values, how results changed over time, what is outside its range, and the text of your reports. Nothing else.
That last line is the quiet one that matters most. DoctanoAI is scoped to your profile: it can see your results and your reports, it will not invent a result that is not in them, and where nothing usable exists it says so instead of guessing. A dedicated timeline with one person's results per profile is a cleaner context than a chat window will ever be.
Why Doctano is the right home for your context
- It takes the reports you already have. No portal connection, no participating provider. If you can photograph it or save it as a PDF, Doctano can read it.
- Cross-lab normalization is the product, not a feature. SGOT and AST land on one series, and mg/dL converts to mmol/L so the trend is real.
- Every value can be checked. The original report is one tap from any result, so a misread gets caught instead of memorized.
- Each lab's own range is kept per result. Classification uses the range printed on the report that ran the test, not a global table.
- One context per person. Family profiles keep each person's timeline separate, so answers about one person come only from that person's results.
- Free, and private by design. Every feature is on the free plan, nothing is public, everything behind sign-in is private to your account, and you can delete your account and all linked data from the app.
Where Doctano falls short, in plain terms: it does not diagnose, and DoctanoAI does not replace a doctor. Data export is not available yet. Extraction can misread a dense table, which is exactly why every value stays linked to its source. And context makes AI answers relevant; it does not make them medical advice.
Ten minutes to your first real answer
- Create your profile at app.doctano.com. Google Sign-In, free, no card.
- Upload your oldest report and your newest one, as PDFs or photos.
- Open each report and spot-check a few values against the original. The document is one tap from any result.
- Ask DoctanoAI what changed between the two reports, and which results are outside their reference range.
- Take the summary to your next appointment. The AI organizes the question; your doctor interprets it.