How our AI works

• Every answer cites the documents it used. If nothing in your records supports an answer, it says so.
• Questions asking for a diagnosis, a dose or a treatment decision get facts from your own records plus “please discuss this with your doctor”.
• Values the reader is less than 80% sure about wait for your Confirm or Edit before they show in Trends.
• We log each AI result with counts and labels only — never names, document text or your values — so we can measure accuracy and switch a model off instantly if it misbehaves.

What it does: Reads the reports, prescriptions and photos you upload and files the values, medicines, conditions and allergies it finds.

Not for: Diagnosing anything or reading documents of people outside your household.

If it's switched off: Uploads are kept safely but not read; you can retry once reading is back on.

Feedback: Every value has Confirm and Edit. Values under 80% confidence wait for your check before they appear in Trends.

What it does: Writes a short plain-language summary of each report.

Not for: Telling you whether results are good or bad news.

If it's switched off: Reports are filed with their title only, no summary.

Feedback: Use “Useful?” under the filed report card in chat.

What it does: Answers questions about your own records and cites the documents it used.

Not for: Diagnosis, medicine doses or treatment decisions — it will point you to your doctor.

If it's switched off: The assistant lists matching documents from your vault instead of writing an answer.

Feedback: Use “Was this right?” under any answer.

What it does: Spots questions asking for a diagnosis, a dose or a treatment decision, so answers stick to your records and send you to your doctor.

Not for: Medical triage.

If it's switched off: Every question is treated as a clinical request (the safest setting).

Feedback: Use “Was this right?” under the answer.

What it does: Lines up your readings over time against the ranges printed on your reports.

Not for: Predicting future results.

If it's switched off: Only confirmed values are shown.

Feedback: Correct any value from the original report in your Vault.

What it does: Points out patterns across your values, medicines and symptoms that may be worth discussing with your doctor.

Not for: Diagnosis or treatment advice.

If it's switched off: The Insights page shows your records without derived insights.

Feedback: Tell us what looked wrong through the Contact page.

What it does: After a new report, posts a short card listing values that changed since the last one.

Not for: Clinical interpretation.

If it's switched off: No comparison card is posted.

Feedback: Tell us what looked wrong through the Contact page.

What it does: Checks symptoms and readings from your other Atriverse apps against your medicines and reports.

Not for: Deciding whether to change a medicine.

If it's switched off: Other-app events are stored but not cross-checked.

Feedback: Tell us what looked wrong through the Contact page.

Use “Was this right?” under any answer, Confirm or Edit on any value, or write to us via the Contact page. Corrections are reviewed and used to measure accuracy.