Apple Health's State of Mind feature is a structured reflection, not a clinical assessment. A person can log how they feel right now as a momentary emotion or describe how they felt overall as a daily mood, then optionally add words and context about what influenced the entry.

That structure is useful for a personal timeline, but it does not turn a mood log into a diagnosis or a causal finding. A careful AI analysis preserves the entry kind, valence, labels, associations, timestamp, and missing context before describing any pattern.

At a glance

What to keep in view

  • Daily Mood and Momentary Emotion are different HealthKit entry kinds and should not be pooled without a declared method.
  • Valence, labels, and associations describe what a person recorded; they do not establish a clinical state or cause.
  • Lifestyle associations are context for reflection, not evidence that sleep, exercise, or daylight caused a mood change.
  • MCP can carry a minimized, user-approved summary to an AI client, but it does not grant HealthKit access or provide mental-health assessment.

Two kinds of State of Mind entry

Apple's Health guide distinguishes a momentary emotion—how a person feels right now—from a daily mood—how they have felt overall that day. Both can be logged with a valence slider, optional descriptive words, and optional information about what had the biggest impact.

The distinction changes the denominator for an analysis. Several momentary entries in one day do not equal several daily moods, and a missing momentary entry does not mean the day was emotionally neutral. Keep the kind and timestamp attached to every record.

  • Momentary Emotion: a time-specific reflection.
  • Daily Mood: an overall reflection for the day.
  • No entry: an unknown state, not a neutral score.

What HKStateOfMind stores

HealthKit's HKStateOfMind sample exposes the entry kind, a valence value and classification, labels, and associations. Those fields represent the person's recorded reflection and selected context. They are not sensor measurements and are not an objective monitor of mental health.

An export or AI payload should preserve the sample date, entry kind, valence, labels, associations, source, and any available metadata. Do not reduce a nuanced set of labels to a single wellness score without documenting the transformation.

Associations are prompts for reflection, not causes

Health lets a person review factors they associated with an entry and compare it with lifestyle factors such as exercise, sleep, time in daylight, and mindful minutes. Apple presents these comparisons as a way to learn about personal patterns.

A co-occurring change is not proof of causality. Fewer daylight minutes and a lower valence on the same day may reflect a third factor, selection bias, or ordinary variation. A trustworthy summary says the records co-occurred in the selected window and names what remains unobserved.

A synthetic timeline

The following values are invented and contain no real wellbeing data.

  • Monday 08:10 — Momentary Emotion, slightly pleasant, labels selected; no association recorded.
  • Monday 20:30 — Daily Mood, neutral, with sleep selected as a context factor.
  • Tuesday — no State of Mind entry; the person's mood is unknown for this dataset.
  • Defensible summary: two different entry kinds were recorded on Monday; Tuesday has no observation.

How an AI should summarize the records

Before asking a model for a summary, include the selected date range, time zone, entry kind, timestamp, valence and classification, labels, associations, source, and missing days. State whether repeated entries were kept, grouped, or excluded.

The model may describe distributions, changes in the person's own logged reflections, and questions worth exploring. It should not infer depression, anxiety, resilience, treatment response, or a cause from State of Mind entries alone, and it should never invent a population normal range.

MCP may be an optional third-party interface for a minimized, user-approved summary. It is not an Apple protocol, does not grant HealthKit permission, and should not silently transmit free-text context or a lifetime mood history.

Privacy and mental-health boundaries

Mood labels, free-text context, and associations can reveal highly sensitive information. Request only the dates and fields needed for the user's question. Prefer local aggregation, minimize free text, disclose external model processing, and make deletion and revocation visible.

This article is a data-model and interpretation guide. It does not diagnose a mental-health condition, assess suicide risk, recommend treatment, or replace a qualified professional. Urgent safety concerns require appropriate emergency or clinical support.

Use boundary

Information, not medical advice

This article explains data and research methods. It does not diagnose a condition, prescribe treatment, establish a universal normal range, or replace qualified professional care. If symptoms or a medical decision concern you, use an appropriate clinical service.