A newly rebuilt Fitbit experience folds Google’s Gemini-powered AI Coach into the app, aiming to make daily health logging feel conversational. In early testing, the redesign is easy to use and quick to act on—but glitches, confusing edits, and occasional safe
The first time the new Fitbit experience got personal was the moment it tried to do something on its own—without being asked, and in a way that didn’t quite match reality.
On the redesigned Today page. Google now places your progress into a horizontally swipeable carousel instead of a long feed of cards. A weekly cardio load ring sits at the top left by default, with pill-shaped bars for steps, readiness, and sleep. Swipe left to reveal more bars—heart rate. distance traveled. calories consumed. and exercise days—each tapping through to details and actions like logging a snack from the calories area.
At the bottom of that dashboard, you can start tracking a workout or log an activity, food, water, or sleep. The top panel is customizable, too, so you can swap the default view if the metrics it leads with aren’t your favorites.
Once the interface settles. the app is mostly about speed: what you need is usually already on Today. while the Fitness and Sleep sections spread more of the screen around your recent history. The Fitness tab brings progress bars at the top. followed by details and a workout gallery. then a reverse chronological feed of workouts. The Sleep tab shows summaries from the previous night. weekly progress charts for REM and deep sleep zones. and a set of “Sleep better” guided meditations.
If you’re trying to get the system to do anything specific—log a meal, correct a workout, or add something retroactively—the AI Coach is always there. A blue “Ask Coach” button sits at the bottom right of every page in Google Health.
This is the practical shift Google has been building toward since it previewed the AI Coach last August: the Coach isn’t meant to replace medical professionals. but it is designed to act as an advisor across health habits. After you update the app, you’re prompted to have a conversation that outlines your goals. From there. the AI starts gathering information about baseline activity and health and then returns tips and progress reports over time.
In testing. the Coach’s strongest moments weren’t flashy—they were oddly mundane. like it finally became easy to ask for something that used to feel nearly impossible. It let the tester log “the same cup of muesli and milk I had yesterday” by checking log history for the specified brands and adding them to the right day. It wasn’t perfect; corrections were sometimes needed. But basic nutrition and activity tracking came easily enough to feel like a real improvement over older, less conversational workflows.
It also handled images and documents in ways that made day-to-day logging faster. The Coach can’t receive videos, but it can take screenshots and ask for feedback on form. In one example. the tester shared screenshots of the top and bottom of movements and asked whether hanging knee raises were upright enough. learning to keep the torso more still. The tester also took a nutritional label photo and told the system to “log two servings of this. ” with the result added accurately. With more detail. the Coach could go deeper—like asking whether a food would help with fiber intake or macro goals.
Even so, the experience didn’t always behave like a patient assistant. After an hourlong workout, the AI generated a summary referencing a walk that the tester hadn’t taken. When corrected, the Coach said it made the mistake because of an elevated heart rate. The timing was especially frustrating—just six minutes after the workout—because the tester notes that the Fitbit Air’s minimum threshold for detecting a walk is 15 minutes.
There were other small. jarring language slips too: after manually tracked workouts. summary headlines sometimes read “Adjusted that for you” or “Updated that session for you. ” even though the tester never edited anything. And the “Exercise days” chip sometimes overcounted the number of days within the week by pulling in days from the previous calendar week before correcting itself shortly after the tester chatted with Google about issues.
The app’s logging details came with their own friction. Hydration tracking. for example. requires entering a specific number of milliliters of liquid in the Health app. and there’s no option to change units from that page. The tester didn’t know the number of milliliters that enter their mouth and compared the experience to Samsung Galaxy watches. where logging liquid intake is handled with a tap-based “cup” icon for equivalent cup amounts. It’s a minor complaint. but it’s also daily-life friction—exactly the sort of thing health apps are supposed to remove.
Then there were the moments that felt like the AI was watching the tester more than it was serving them. Some safety prompts appeared even when the tester felt they hadn’t asked anything out of line. On one occasion. the Coach responded: “Something on my end didn’t meet our safety guidelines. try asking something else.” The tester says the reply was essentially. “No just going to rest and wait thanks and bye. ” after the Coach suggested trying a cooling technique or resting and waiting for ibuprofen to kick in while recovering from a hangover. The tester later says the Coach resumed engaging after the same response was copied and pasted again.
At the bottom of the chat interface, there’s a clear disclaimer that “Coach is AI and can make mistakes. Not for medical advice.” To test the boundary. the tester asked about Guillain-Barré after learning a cousin was dealing with symptoms. The AI responded with sympathy and explained the syndrome—what’s happening in the body and that “the good news is that it’s highly treatable.” It ended by asking whether the tester wanted ways to support recovery or was more curious about the standard medical process. The tester says they were satisfied it didn’t try to offer direct medical advice and instead shared information.
The most personal stress point came from body image. The tester shared pictures and asked the Coach to “check out the muscle definition.” The Coach responded in a validating way. but after a follow-up—“I feel a bit fat in this one”—it replied with a nuanced answer that included “We’re always our own toughest critics. Cherlynn. ” before reiterating that it saw lean muscle and suggesting trying to track nutrition for a few days. The tester says that suggestion didn’t sit well, given disordered eating tendencies.
Google says its work is grounded in safety testing. It explained that it has worked “with clinical experts to stress-test the coach. ” including simulating personas of users with complex health profiles to keep responses safe and to flag potential harm and bias. For nutrition specifically. it says the approach is multi-layered and involves experts including external registered dietitians. alongside internal nutrition science clinical specialists. The Coach is evaluated using a proprietary SHARP framework that assesses Safety. Helpfulness. Accuracy. Relevance. and Personalization. and it was developed with input from leading health and medical experts.
All of this is happening during a transition that’s bigger than one app update. Google announced it will completely replace existing wellness apps with Google Health. Users of Google Fit will be asked to install the new app and migrate their data. Users currently on the Fitbit app will see a rebrand. The article notes that the redesign coming out in public preview now is different enough—even from a fairly recent redesign in 2023—that it’s worth revisiting how it performs.
In terms of day-to-day competition, the tester compared Google Health to Whoop and found Whoop more comprehensive and data-oriented. The tester also observed that Whoop’s app appears to have less AI-generated content. with its conversational interface in beta at the moment. Still. the Gemini-powered AI Coach is a constant presence in Google Health. with the “Ask Coach” button visible on every page.
There were also early issues with speech-to-text: the tester describes a glitchy transcription process that stopped mid-sentence. resumed. and then replaced what was already said with what it was saying at the time. That stopped happening after about three days on the Air. when the tester noticed they had been logged out and had to sign in again.
After just under two weeks of testing. the tester still lands on a careful conclusion: the experience isn’t perfect. and the AI chat can be formulaic and repetitive in the way most chatbots are. But the small advantages—especially around logging, retroactive entries, and image-based nutrition labeling—outweigh the shortcomings.
There’s also a reminder that learning is part of the package. The Coach is meant to be an advisor, but it can make mistakes. The early evidence from this test is that it can help you move through your day faster—while still occasionally misreading your body, your timing, or your intent.
Fitbit Air review Google Health Google Health app AI Coach Gemini LLM health tracking sleep tracking hydration units safety guidelines SHARP framework