Apple Health Wrapped 2026

What's in it

Totals for steps and energy are Apple Health's own daily totals, so they match the Health app. VO2 max only appears if your watch records it. Workout types HealthAPI doesn't name, such as tennis or skiing, are grouped as other; the named ones are on the metrics page.

Before you start

• Have the whole year synced. HealthAPI imports a year of history by default. If you installed it recently, check that Settings, Sync Settings, Import History says your history is up to date.
• Connect your AI. Claude on every plan (guide), ChatGPT on Plus or higher (guide), approved on your iPhone.
• Make it in December. The year-long lookups count back 365 days from today, so on December 31 they cover the whole year. After that, see the January answer below.
• Two sleep trackers are fine. If an Apple Watch and a ring or band both record your nights, each night still counts once (Oura, WHOOP, Garmin).

The prompt

Paste it into a new chat with HealthAPI connected. Naming the lookups keeps the AI from reading your year day by day, which would take hundreds of requests.

Make my Apple Health Wrapped 2026 from HealthAPI, using as few requests as you can:
1. get_health_overview once, to see what I have.
2. get_trend with days 365 for stepCount, distanceWalkingRunning, activeEnergyBurned, restingHeartRate and vo2Max. Keep only 2026 dates.
3. get_workouts from 2026-01-01 to 2026-12-31, limit 100. Only if it says the list was cut short, fetch it a quarter at a time.
4. get_sleep for 2026, in four calls of up to 92 nights.
Don't fetch day by day or week by week.

Then show:
- total steps, distance and active energy
- workouts by type, with count and hours
- average sleep per night, by month
- my best month for steps and my most active day
- how my resting heart rate and VO2 max changed from January to December
Skip anything I have no data for, and list the totals you used at the end.

In Claude: an artifact

Add this line to the prompt to get a page you can keep:

Build it as an artifact: one page, a card per stat and a bar chart of steps by month, with the numbers written into the page.

Artifacts work on every Claude plan. If Claude answers in plain text instead, check that code execution and file creation is on under Settings, Capabilities.

In ChatGPT: right in the chat

OpenAI took canvas out of ChatGPT's current models in May 2026, so ask for the recap in the chat:

Show it here in the chat: a short summary, then a table for each section.

For a page to keep, add “Then give me the whole recap as one HTML file in a code block”, save it as wrapped.html and open it in your browser.

How many requests it takes

That fits a day on any plan: 25 requests a day during the preview, 100 on Pro and 1,000 on Developer, with room for follow-up questions. Requests reset at midnight UTC. If you see the AI fetching one day or one week at a time, stop it and paste the prompt again.

Before you share it

A Wrapped is made to be shared, but this one is your health record. Read it first, and decide what you're happy for others to see.

• Claude. The numbers Claude writes into an artifact go wherever the artifact goes. On Pro and Max you can set it to Anyone with the link; set it back to Only you to stop sharing.
• ChatGPT. Anyone with a link from a personal ChatGPT account can view the shared chat, and OpenAI advises against putting health information in shared content. Delete links under Settings, Data controls, Shared links.
• Make a share version. Ask for “a second version with only steps, distance and workouts, no sleep, heart rate or dates”, or post a screenshot of the cards you're happy to show.
• Your account stays yours. A shared link shows what's on the page and nothing more. Only the AI apps you approved on your iPhone can read your HealthAPI data, and only to read it.

HealthAPI and the AI apps aren't medical devices. A change in resting heart rate or VO2 max over a year is a pattern to notice, not a diagnosis; talk to a doctor about anything that worries you.

For developers: the same recap over REST

With your API key the recap takes about 60 requests: five trends, one call for every workout, and 52 or 53 weekly summaries for sleep, since the REST API has no per-night range. That fits Pro's 100 a day. Setup and the key are in Apple Health data in Python.

import os
from datetime import date

import pandas as pd
import requests

API = "https://healthapi.app/api/v1"
HEADERS = {"Authorization": f"Bearer {os.environ['HEALTHAPI_KEY']}"}
YEAR = 2026


def get(path, **params):
    r = requests.get(API + path, headers=HEADERS, params=params, timeout=60)
    r.raise_for_status()
    return r.json()


def year_of(metric):
    """One value per day of YEAR: the total for steps and energy, the average otherwise."""
    t = get("/query/trends", type=metric, days=365)
    col = "total" if t["aggregation"] == "sum" else "avg"
    s = pd.Series({d["date"]: d[col] for d in t["data"]}, dtype=float)
    s.index = pd.to_datetime(s.index)
    return s[s.index.year == YEAR].sort_index()


# 5 requests
steps, km = year_of("stepCount"), year_of("distanceWalkingRunning") / 1000
kcal, rhr = year_of("activeEnergyBurned"), year_of("restingHeartRate")
vo2 = year_of("vo2Max")
print(f"{steps.sum():,.0f} steps, {km.sum():,.0f} km, {kcal.sum():,.0f} kcal")
print("Best month:", steps.resample("MS").sum().idxmax().strftime("%B"))
print("Most active day:", kcal.idxmax().date())
if len(rhr):
    print("Resting HR:", round(rhr.iloc[:30].mean()), "->", round(rhr.iloc[-30:].mean()))
if len(vo2):
    print("VO2 max:", round(vo2.iloc[0], 1), "->", round(vo2.iloc[-1], 1))

# 1 request: every workout of the year
w = pd.DataFrame(get("/health/workouts", start_date=f"{YEAR}-01-01",
                     end_date=f"{YEAR}-12-31", limit=1000))
if len(w):
    by_type = w.groupby("activity_type").agg(
        workouts=("duration_seconds", "size"), hours=("duration_seconds", "sum"))
    by_type["hours"] = (by_type["hours"] / 3600).round(1)
    print(by_type.sort_values("workouts", ascending=False))

# 52 or 53 requests: sleep per night, from the weekly summaries
nights = []
for week in range(1, date(YEAR, 12, 28).isocalendar()[1] + 1):
    summary = get("/health/weekly-summary", week=f"{YEAR}-W{week:02d}")
    for day in summary["daily_breakdown"]:
        if day["date"].startswith(str(YEAR)) and day["sleep"]["total_hours"] > 0:
            nights.append(day["sleep"]["total_hours"])
if nights:
    print(f"Average sleep: {sum(nights) / len(nights):.1f} h over {len(nights)} nights")

To compute everything offline instead, download the JSON export, which doesn't count toward your requests. Run the script in December for the same reason as the prompt: days=365 counts back from today.

Questions