How your health data is used

Your readings stay on this phone. Your account — sign-in, goals, preferences — is stored on a server. Nothing about your body is sent anywhere unless you turn on AI narration, and that is off unless you turn it on.

Last updated 2026-08-12

What is read

BiomarkersIQ reads from Apple Health only the categories you chose during setup, and you can change them at any time in the Health app.

Access is read-only. The app never writes to Apple Health and does not ask for permission to. Apple shows the words "and update" on its own permission sheet for every app; that wording is fixed by iOS and is not something an app controls.

Where it is kept

Every reading, every daily value computed from it, and the record of which samples went into that value are stored on this device and nowhere else.

The account layer is separate and holds no health data: your sign-in identity, your goals, your preferences, and which tier your account is on — everyone is on the free tier today, because there is nothing to buy yet. That lives on a server so you can sign in again after reinstalling.

Signing in does not upload health data. It never has.

What leaves the device, and only if you allow it

The app can optionally send a description of a week or month that has finished to Anthropic, the company behind the Claude models, so the report reads more naturally than the app can write it alone. This is off unless you turn it on, and turning it off stops it immediately.

The first part of what is sent is the findings you were shown: the kind of each finding, which measurements it concerns, the handful of numbers its sentence is allowed to quote, how much of the period was measured, and the period’s dates. A finding the app worked out but did not show you is not sent.

The second part is one row for each measurement you actually recorded, whether or not it produced a finding: its average over the period, your usual figure for it, its highest and lowest days, and how many days of the period it was measured on. This is what lets the model describe what the period was like rather than only what stood out in it. Your usual figure is included deliberately — the app does not print it on every card, and a description of a period without it is a list of averages with no sense of whether any of them was ordinary.

Each finding may also include BiomarkersIQ’s own assessment of the finding, such as its confidence, the strength and direction of the relationship or change, whether the change is considered an improvement or worsening, and calculated differences between the measurements. These limits help keep the AI-generated wording consistent with what BiomarkersIQ has already determined from your data.

To help choose which relationships are worth testing, the model may receive the names and types of available measurements and how many days of data are available for each. No measurement values are sent for this step.

It contains no name, no email, no account identifier, no device identifier, and no series of readings. No date in any of it is more precise than a day.

The paragraph at the top of a weekly or monthly report is written by the model rather than reworded from a finding, from those per-measurement rows, and it is not allowed to state a number. Every other sentence a model touches is a rewording of something the app worked out, checked against the same numbers before you see it.

You can see the exact format of what would be sent before you agree to any of it, on the AI narration screen — a real example, printed by the same code that builds the request. Every sentence in the app records whether the app or a model wrote it.

BiomarkersIQ keeps no copy of what is sent. Anthropic does: they delete it within thirty days, and may keep it for up to two years if their automated safety systems flag it. That is their policy rather than ours, and it is the one part of this the app does not control.

Anthropic does not use data sent through its API to train its models.

If you turn this off, or never turn it on, every sentence in the app is written by the app itself. Nothing is withheld from you for declining.

What the app will not say

BiomarkersIQ does not diagnose, treat, or predict. It describes what your own measurements did, compared against your own history, over periods that have finished.

It does not compare you to population averages, does not produce a score, and does not tell you whether a reading is good or bad — because for most measurements that depends on things the app cannot know.

It is not a medical device and is not a substitute for advice from a clinician.

Deleting it

Deleting the app removes every reading and every finding stored on this device.

You can also delete your account from inside the app, in Profile. That erases your account and everything held with it — your sign-in identity, your goals, your preferences — and wipes every reading, daily value and finding from this device at the same time. It happens straight away, and it cannot be undone.

Your Apple Health data is not touched. BiomarkersIQ only ever reads from it, and everything in the Health app stays exactly as it is.

If you signed in with Apple, deleting your account also withdraws this app’s access to your Apple ID. You will be asked to sign in with Apple once more first so that access can be withdrawn — that sign-in confirms it is you, and nothing from it is kept.

Questions about any of this go to support@biomarkersiq.com.