The five-second answer: if you mainly want step trends, Google Fit's weekly view already wins most of the time
The goal sounded simple: take the steps and activity minutes already recorded on an Android phone, expose them to ChatGPT, and ask things like, “Am I walking more this week?”, “How much did I increase from last week?”, or “If my food intake stays the same, what does this activity trend imply for weight?”
Technically, it works. On Android, you can build this pipeline:
Google Fit → Health Connect → freddy → ChatGPT.
Once connected, ChatGPT can query real values such as steps, distance, estimated calories, and workout duration instead of guessing from a screenshot. [1][2][3]
Then comes the punchline: freddy's currently advertised Free plan is one source and seven days of history. [4] The moment you ask, “Compare May-to-now with the months before May,” the fancy AI analytics machine drives directly into a seven-day brick wall.
For long-term step trends, Google Fit's built-in Day / Week / Month views may simply be the better tool. [5] We built the spaceship and then discovered the stairs were fine.
What were we actually trying to build? A way for Google Fit data to talk to an AI
Google Fit can record steps, distance, activity time, and workouts. What it does not primarily try to be is a free-form analyst that answers arbitrary natural-language questions about your personal trend.
That is where Health Connect matters. Google positions Health Connect as the Android layer that lets health and fitness apps store and share supported data types under user-controlled permissions. [1][2]
The architecture is straightforward:
Google Fit records → Health Connect brokers → freddy translates for AI → ChatGPT analyzes.
Despite the name, this freddy does not patrol a haunted pizza restaurant at night. It is a health-data MCP service. No animatronic bear required. [4][6]
What becomes visible after connecting it? More than just steps
When freddy is allowed to read Health Connect, it can pull the data types you authorize, including steps, distance, calories, workouts, heart rate, sleep, and others—provided some app or device has actually written those records. freddy states that its Health Connect integration is read-only and does not write back to Health Connect. [3]
That means ChatGPT can request recent values by date and compare them directly. A line chart can also be created from the retrieved numbers.
One practical warning: if you ask an AI to “make a line graph,” it may occasionally interpret that as “generate an image.” That is not the same thing. For quantitative tracking, you want a real plotted chart from the actual series—not an inspirational painting about step counts.
Does this make your Google account less secure? Separate account security from health-data privacy
Health Connect permissions are scoped by health-data type and access mode. Apps request permissions such as reading steps or heart rate, and users can grant, deny, or revoke them. [2][7]
So this is not equivalent to giving a third party your Google password.
However, adding a health-data service still adds a privacy and security surface. freddy says it syncs connected health data to its own service so that an authorized AI client can retrieve those metrics later. [6][8]
freddy says stored health metrics and provider credentials are encrypted, and that it does not sell health data or use it to train models. [8] That is useful, but it does not magically make third-party storage risk-free.
The most important operational detail is the personal MCP URL. freddy's Terms explicitly warn that anyone who has that URL can query the user's health data. [6] Treat it more like a secret token than a normal webpage link. Do not post it in screenshots, forums, or social media.
What about sleep quality? Health Connect is a warehouse, not a sleep sensor
Installing Health Connect does not cause an Android phone to suddenly know your deep-sleep and REM percentages.
Health Connect is the exchange layer. A watch, ring, sleep app, or other source must first create sleep records. Only then can another authorized app read them. [2][3]
So it is completely normal for steps to appear while sleep is empty. Google Fit may already be writing activity data, while nothing on the phone is writing sleep records.
If you need to buy another wearable, install another tracker, manage another permission set, and charge another device just to obtain a “sleep score,” you can eventually reach the beautiful state where sleep tracking reduces your sleep. Sometimes the simplest metric—actual sleep duration—is enough.
If food stays the same and walking increases, should body weight drop?
In basic energy-balance terms, sustained higher energy expenditure with genuinely unchanged intake pushes body weight downward over time.
Real life is messier. People may eat slightly more after exercising, or unconsciously reduce non-exercise movement because they are tired. A 2023 review found compensatory decreases in non-exercise physical activity in a substantial share of exercise-training studies. [9]
Short-term body weight also moves with water, glycogen, sodium, and digestive contents. Therefore, “I walked a lot yesterday and the scale did nothing this morning” is not especially informative.
A better approach is to compare multi-week averages: weekly activity against weekly average body weight, rather than treating every morning as a final exam.
The biggest limitation: “compare the last year” meets the seven-day Free plan
The connection worked. The data was real. The AI could query it. That part was genuinely useful.
The limitation appeared when the question became historical. freddy currently advertises the Free tier as one source, seven days of history. [4]
That makes it good for questions about this week. It does not make it a free replacement for long-term fitness-history analysis.
This is not necessarily a flaw in the product. freddy is aimed at people who want an AI to combine multiple data sources—Garmin, Oura, WHOOP, glucose monitors, sleep, HRV, lifting logs, and more. [4]
But if your entire requirement is “show me whether I have been walking more since spring,” the value proposition becomes slightly comic:
Google Fit: shows longer-term activity views for free. freddy Free: lets the AI ask your data questions—but only over seven days.
The product is not useless. The use case is simply much narrower than “AI for anyone who has steps.”
For long-term step comparisons, Google Fit's weekly and monthly views are enough
Google Fit officially supports switching activity views between Day, Week, and Month. [5]
That covers a large number of ordinary questions:
- Am I walking more lately?
- Was this week more active than last week?
- Is my monthly activity rising or falling?
- Is the trend obviously moving upward?
If you want deeper interpretation, send ChatGPT a screenshot or exported slice only when needed. That keeps the long-term archive in the fitness platform and sends only the relevant section to the AI.
Not every useful workflow requires a permanent four-service pipeline.
Who benefits from freddy, and who is fine with Google Fit?
freddy makes sense when
- you use several wearables or health platforms;
- you want sleep, HRV, heart rate, recovery, training load, and activity analyzed together;
- you ask cross-domain questions such as “Why am I unusually tired this week?”;
- you are willing to pay for long-term AI-readable history;
- you actually like MCP and agent-style integrations.
Google Fit is probably enough when
- you mainly care about steps and activity time;
- weekly and monthly trend views answer the question;
- you prefer not to add another health-data processor;
- you do not want another subscription;
- occasional screenshots to ChatGPT are good enough.
If you are in the second group, there is no prize for turning a pedometer into distributed systems engineering.
Final verdict: the AI integration works, but “just use the weekly view” is also correct
The technical result is legitimate: Google Fit → Health Connect → freddy → ChatGPT can work, and it lets an AI read and analyze actual activity metrics. [1][3]
The security model is not “hand over your Google account,” but it does involve syncing health data to a third-party service and carefully protecting the MCP URL. [2][6][8]
Sleep data requires a real sleep-data source; Health Connect alone cannot invent it. [3]
And for long-term step analysis, the Free tier's seven-day history is a serious limitation. [4]
So the clean division is:
Use Google Fit for ordinary day/week/month activity trends. Use freddy when you genuinely need AI to combine multiple health and fitness streams.
We connected the entire modern health-data stack and ended up pressing “Week” in Google Fit.
Technically: success. Operationally: a full lap around the track back to the original app.
Sources
- Google Fit Help — Health Connect in Google Fit support.google.com
- Android Developers — Get started with Health Connect developer.android.com
- freddy — Connect Health Connect (Android) freddy.coach
- freddy — Official product / current Free-plan limits freddy.coach
- Google Fit Help — Find your activity support.google.com
- freddy — Terms of Service freddy.coach
- Android Developers — Read raw data from Health Connect developer.android.com
- freddy — Privacy Policy freddy.coach
- PubMed — Compensatory Responses to Exercise Training As Barriers to Weight Loss (2023 review) pubmed.ncbi.nlm.nih.gov
