AI Weight Loss Coach: Data-Grounded, Not a Fixed Course

An AI weight loss coach is defined by what a diet app cannot do, read your logged history back and reason about it. A diet app is a calculator with a memory problem. You answer a few questions on day one, it hands you a fixed number like 1,200 or 1,500 calories, and that number does not change no matter what your logs show three weeks later. It stores what you type, but it never reads it back. An AI weight loss coach is built the other way around: it reads your logged history in conversation, notices what changed, and adjusts.

Wellness Project's version of this is Casey Mills, an AI nutrition coach you talk to the way you would text a friend. Instead of a template plan issued once, Casey sets a calorie and protein target from your actual logged weight trend and updates it as that trend moves. If your weight stalls for two weeks even though your logged intake looks on target, that is information the coach can see and act on, not something you have to notice yourself and manually recalculate.

The autonomy angle matters as much as the accuracy. There is no forced daily check-in, no shame messaging for a missed log, and no penalty for logging less on a busy week. You set the pace; the coach reasons about whatever data you give it, and gives you a straight answer when you ask, without guilt-tripping copy wrapped around it.

The specific data involved is your food log entries, weight history, steps, and workouts. In Wellness Project these are not siloed screens you have to mentally cross-reference; they live in one unified context that Casey Mills, and Claude or ChatGPT over MCP, can query together. Ask "why has my weight stalled" and the coach can look at all four signals in the same pass, not just the calorie count.

That matters because a real plateau usually shows up as a pattern across signals, not a single number. Calories logged roughly on target, but steps down 2,000 a day from your baseline and weight flat for ten days, is a different answer than calories creeping up while steps stay constant. A generic calculator has no way to see that combination; it only ever had the one number you typed in on day one.

A deficit of roughly 300 to 750 calories a day below maintenance is a commonly cited sustainable range, and where in that range you land depends on your starting body fat percentage, how fast you want to move, and how much you care about preserving lean mass along the way. Larger deficits produce faster loss on the scale but increase the risk of losing muscle alongside fat, which shows up later as a slower metabolism and a harder time keeping the weight off. If you track body composition with a connected scale, Casey Mills can read that trend the same way it reads bodyweight; see how Withings body composition data feeds into a weight loss plan.

Protein does the heavy lifting for protecting that lean mass during a deficit, which is why Casey Mills sets a protein target relative to your bodyweight rather than a flat gram count that is the same for a 130-pound person and a 220-pound person. A higher protein intake during a calorie deficit is one of the more consistent levers for keeping strength and muscle while the scale moves down. Calories and protein alone only set the stage for that; pairing the protein target with a resistance training plan that gives your body a reason to keep the muscle it has is what actually protects lean mass through a cut.

None of this is issued once and left alone. As new weight, food, and activity data lands, the AI recalculates rather than holding you to a one-time meal plan. That is the practical difference between a coach and a course: a course ends on a schedule someone else set, a coach keeps reasoning about your data for as long as you keep logging.

There are three ways in, and none of them require a wearable to start. Use the app directly on iOS, Android, or web, sign in with Apple or Google, and start logging meals and weight in plain language. Or ask through Claude, which reads your Wellness Project history over MCP once connected. Or ask through ChatGPT the same way, if that is where you already spend your time.

Wellness Project is free during early access across all three surfaces. Connecting a wearable, Apple Health, Fitbit, Oura, or Google Health Connect, reduces how much you have to type by hand since weight, steps, and activity sync automatically, but it is not required to get a working calorie target from day one. You can start with nothing but a food log and a scale reading and add a device later.