November 13, 2026 · 9 min read · Sugam Budhraja

Health Data Permissions in 2026: iOS 27 Lets Users Share Only 30 Days, and iPhone Apps Can't Tell Who Said No

A new user taps Allow and you get a month of data instead of two years. Was it iOS 27's new 30-day option, a new phone, or a quiet no? What Apple Health and Health Connect let your app see after the prompt, how many people say yes, and how to design the ask so fewer of them say no.

A new user taps Allow on your Apple Health prompt. You ask for two years of steps and get back a month.

Maybe they chose the new 30-day option in iOS 27. Maybe they bought their iPhone last month. Maybe they said no to steps and yes to sleep. Your app can tell exactly one of those apart, and Apple designed it that way on purpose.

The moment of asking is where health apps lose the most data, and in 2026 both platforms changed the terms of that moment. This is what your app can and cannot see after the prompt, how many people say yes, and how to design the ask.


What changed in health data permissions in 2026?

Apple added a history choice. In iOS 27, after people pick which data types to share, a second screen asks how much history to give your app: “a recent limited window or their full history” [1]. Developers report the options read “Past 30 Days and Future Data” and “All Recorded Data and Future Data”, and that Allow stays disabled until one is chosen [2]. Apple added an API, getEarliestAuthorizedSampleDate(for:completion:), to tell you when access is limited [3].

Android already worked this way. Health Connect reads data “for up to 30 days prior to the time permission was granted” unless your app also requests the separate history permission, PERMISSION_READ_HEALTH_DATA_HISTORY [4]. Background reads need their own permission too.

So on both platforms, a user can now hand you a month of history on day one. The difference is how much your app is allowed to know about it.

What can your app see after the permission prompt?

On iPhone, three very different users look identical to your app. On Android you can see each state, but you get two chances to ask. Sources: Apple HealthKit documentation; Android Health Connect guides.

On iPhone, a denial looks like an empty database. Apple’s documentation is explicit: “your app cannot determine whether or not a user has granted permission to read data. If you are not given permission, it simply appears as if there is no data of the requested type” [5]. The only read state you can positively identify is the new limited window, and even then Apple says to treat anything earlier as unknown rather than absent [3]. Full access and denied access return the same answer, by design.

You also get one shot per data type. If the user has already decided on every type you request, HealthKit “calls the completion without prompting the user” [6]. Asking again does nothing; only new types trigger the sheet.

On Android, you can see the state but lose the ability to ask. Health Connect tells you which permissions are granted, and you should check every time because users can revoke them at any point [4]. But if a user cancels the request twice in a row, Google’s guidelines say to show an explanation screen, after which “users need to re-enable permissions from within the Health Connect Settings menu” [7]. You get two attempts.

How many users say yes to sharing health data?

Apple and Google do not publish grant rates. The best public evidence comes from research apps, and it shows where the loss happens.

The loss happens at the ask. Once people said yes, almost all kept sharing. Data: Funk et al., PLOS ONE, 2025.

In a study of incoming US medical interns, 59.7% of 1,164 iPhone users allowed at least one of five phone sensor data types during onboarding. Two months later, 94% of those who opted in were still sharing [8]. Opt-in was highest for the least sensitive data (ambient light) and lowest for the most sensitive (frequently visited locations). It was also lower among Asian interns (53.8%) and interns from groups underrepresented in medicine (51.6%) than among White interns (67.0%) [8].

These were people who had already joined a study, so a general consumer app may see different rates. The shape is the useful part: permission is decided once, at the start, and it mostly sticks. That makes the ask the highest-leverage screen in your onboarding.

Does 30 days of health history matter?

Less than it sounds for some features, more for others.

For baselines, a month is roughly enough. Most personal baselines settle within two to four weeks, as we found in how many days a wearable needs to learn your baseline. A user who grants 30 days can get personalised insight from day one.

For anything longer, it is a blank. Year-over-year comparisons, seasonal patterns, personal records and long-term trends such as VO2 max over months all start empty. Developers worry the result will look like a bug: empty charts that users blame on the app, not on a choice they made on day one [2].

The window grows forward. Future data keeps flowing under either option, so a user who grants 30 days and stays for a year ends up with a year of data in your app [2]. The cost falls on the first months, which is exactly when most health app users churn.

How should you ask for health data permissions?

  1. Ask when a feature needs the data, not at launch. Apple’s guidelines: request access “only when you need it”, because a request tied to the current context helps people understand why [9].
  2. Explain before the system sheet, without imitating it. A short screen saying what each data type will do for the user is allowed and recommended. Apple only asks you to avoid “custom screens that replicate the standard permission screen’s behavior or content” [9]. Call it Apple Health, not HealthKit, in anything users read [9].
  3. Ask for less, then more. Request the types your first feature uses. You do not need to request every type at once [1], and on iPhone only new types will prompt again anyway [6].
  4. Prepare people for the history question on iOS 27. If your product depends on trends, say so before the sheet: “Choose All Recorded Data so we can show your trends from day one.” Users must pick an option to continue, so a sentence of context is the difference between a guess and a choice [2].
  5. Detect limited access and design for it. Call getEarliestAuthorizedSampleDate after authorisation, pass its date into your queries, and treat anything earlier as unknown [3]. Make sure trends and anomaly detection work from partial data, and tell users when full access would improve them [1].
  6. Design the empty state as a permissions state. On iPhone, “no data” may mean denied, limited or genuinely empty. Show a gentle path to Settings › Privacy & Security › Health rather than an empty chart. Developers report there is no public link straight to that screen, so describe the path [2].
  7. On Android, count cancels. Check granted permissions before every read, track whether a request has been cancelled twice, and switch to the Settings explanation instead of re-requesting [4][7]. Ask for history and background access separately, at the moment a feature needs them.
  8. Request only what you can justify. Both stores reject apps that ask for health data they do not use, as we covered in App Store rejections for HealthKit and Play Store rejections for Health Connect.

What should you measure?

The number most teams track, the share of users who tap Allow, is the one that matters least, because on iPhone you cannot see it reliably.

  • Effective grant rate: the share of new users whose data for each type actually arrives within 24 to 72 hours. This counts denials, empty stores and broken flows together, which is what your product feels.
  • Limited history rate: on iOS 27, the share of users with an earliest authorised date, by type.
  • Android cancels: the share of users who cancel once, and twice.
  • Revocations: users whose data stops arriving after it started, which is separate from churn.
Where Sahha sits. Sahha’s SDK reports a sensor status of pending, unavailable, disabled or enabled through getSensorStatus(), so an app can tell whether it still needs to ask and when to show onboarding again or a recovery banner. Its documentation recommends a short explanation screen before enableSensors() triggers the system prompt [10].

Health data permission used to be a single tap at the end of onboarding. In 2026 it is a two-step decision on iPhone and a two-strike rule on Android, and on iPhone your app cannot even see the answer. The teams that do well will treat the ask as a product feature: timed to a moment of value, explained in plain words, and measured by the data that actually arrives.

References

  1. Apple Developer. Authorizing access to health data. HealthKit documentation. https://developer.apple.com/documentation/healthkit/authorizing-access-to-health-data
  2. Apple Developer Forums (2026). iOS 27 Health permissions: a reckoning is coming. https://developer.apple.com/forums/thread/842499
  3. Apple Developer. getEarliestAuthorizedSampleDate(for:completion:). HealthKit documentation, iOS 27.0 and later. https://developer.apple.com/documentation/healthkit/hkhealthstore/getearliestauthorizedsampledate(for:completion:)
  4. Android Developers. Get started with Health Connect. https://developer.android.com/health-and-fitness/guides/health-connect/develop/get-started
  5. Apple Developer. authorizationStatus(for:). HealthKit documentation. https://developer.apple.com/documentation/healthkit/hkhealthstore/authorizationstatus(for:)
  6. Apple Developer. requestAuthorization(toShare:read:completion:). HealthKit documentation. https://developer.apple.com/documentation/healthkit/hkhealthstore/requestauthorization(toshare:read:completion:)
  7. Android Developers. Permissions and data access. Health Connect UI guidelines. https://developer.android.com/health-and-fitness/health-connect/ui/permissions
  8. Funk, C., Zhao, Z., Horwitz, A. G., Fang, Y., Pereira-Lima, K., Kheterpal, V., Sen, S., and Frank, E. (2025). Feasibility and acceptability of collecting passive phone usage and sensor data via Apple SensorKit. PLOS ONE. https://doi.org/10.1371/journal.pone.0329707
  9. Apple. HealthKit. Human Interface Guidelines. https://developer.apple.com/design/human-interface-guidelines/healthkit
  10. Sahha. Data collection. Sahha Docs. https://sahha.ai/docs/connect/sdk/collection/

Related