Here’s a bold statement: AI didn’t write my workout plan, but it transformed my cycling performance. And this is the part most people miss—it’s not about letting AI take the wheel; it’s about using it as a smart co-pilot. Let me explain.
I’ve been experimenting with AI, specifically Google Gemini Pro, for two months, and the results are eye-opening. I’ve lost weight, gained strength, and feel more energized—all without AI dictating my workouts. But here’s where it gets controversial: while many fail using AI for training plans, I succeeded by focusing on smaller, specific tasks like nutrition and recovery analysis. Why? Because I stayed in control, using AI as a tool, not a trainer.
The Setup
I chose Google Gemini Pro because I’m already in the Google ecosystem, and its paid version offers better models and fewer limitations. Pairing it with Garmin Connect and hardware provided the data depth needed for meaningful insights. This combo wasn’t just about convenience—it was about precision.
How It Started
My journey began with a health concern: low energy and motivation, reminiscent of past struggles with an eating disorder. Instead of asking AI for a training plan, I started with a specific question: ‘Should I get tested for low testosterone after a big ride or on a rest day?’ Gemini’s follow-up questions led to a revelation: I was under-eating and over-training, especially during the abrupt shift from summer to winter cycling.
How It Progressed
Gemini became my daily nutritionist and recovery coach. Every morning, I’d share Garmin data, and it would tweak my meals—like suggesting more protein or carbs based on my training. For instance, it helped me design a better breakfast of overnight oats with Greek yogurt, tailored to my evening rides. It even analyzed a Christmas platter, advising what to eat more of and what to avoid. But I always led the conversation, asking for tweaks, not full plans.
During rides, I’d ask specific questions like, ‘Should I do a VO2 max workout today or wait?’ Gemini interpreted Garmin’s data, offering insights without dictating my training. The result? I lost weight unintentionally, drink more water, and recover better. Most importantly, I’m having fun on the bike again.
AI as My Directeur Sportif
I even used Gemini during a Zwift race. It analyzed the course, suggested pacing, and advised when to attack. Though I finished near the back, Gemini’s post-race analysis was a game-changer. It explained why I lost (raw power vs. weight) and highlighted a physiological victory: I held 96% of my FTP for 40 minutes—a massive threshold breakthrough. It also synced with Xert to validate my performance, boosting my confidence.
AI’s Limitations
But AI isn’t perfect. Here’s the controversial part: you need to be the expert, not the student. AI is great for pattern recognition and guidance, but it can’t replace human judgment. For instance, it got fixated on my cadence and a specific bike, despite my requests to drop the topic. Interaction is another issue—typing works, but voice commands are clunky. And don’t expect it to create a six-month training plan; tools like TrainerRoad or a human coach are better for that.
Final Thoughts
AI’s strength lies in its ability to enhance, not replace, your expertise. Break tasks into small, specific questions, and always verify its output. If you’re curious about AI’s potential, start with something simple, like asking what to eat before a workout. But remember, the real power is in the partnership—not the technology itself. What do you think? Is AI a game-changer or just a fancy calculator? Let’s discuss in the comments!
