Can AI Create a Meaningful Yoga Practice? A Doctor’s Capture
AI tools like ChatGPT have become ubiquitous, offering assistance with research, learning, and even creative endeavors. But as these tools gain popularity, a critical question arises: what tasks are best left to human expertise? Whereas ChatGPT can help plan a vacation, can it curate a practice as deeply embodied as yoga?
We tasked ChatGPT with creating a 60-minute vinyasa yoga sequence with Crow pose as the peak. We then asked experienced yoga teachers to evaluate the resulting practice, revealing both the potential and limitations of AI in this context.
What ChatGPT Got Right
ChatGPT demonstrated a basic understanding of yoga sequencing, recognizing the need for a warm-up. “It understood that the wrists needed to be warmed up, the heat needed to be built,” noted yoga teacher Taylor Lorenz. The inclusion of a five-minute Savasana at the end was also appreciated, as it’s a crucial integration step often overlooked.
Interestingly, ChatGPT suggested pulsing in Goddess Pose to prepare for Crow, a unique cue that Lorenz found effective for warming up the inner thighs. The bot also encouraged students to explore the shape of Crow with their feet on the mat before attempting the full balance, a helpful approach for building confidence.
One unexpected and well-received transition was moving from Revolved Crescent Lunge to Goddess Pose facing the opposite side of the mat. Yoga Journal editor-in-chief Renee Marie Schettler described it as “intuitive and lovely.”
Where the AI Practice Fell Short
Despite some positive aspects, the AI-generated practice lacked the creativity and physicality needed for a truly effective yoga class. Testers found it repetitive and, uninspiring.
The sequence heavily relied on Sun Salutations—22 minutes dedicated to Sun A and B, plus three rounds of Half Sun Salutations. While repetition isn’t uncommon, it felt forced and potentially exhausting before the challenging Crow pose.
Transitions between poses like Warrior 1 to Warrior 2 and Goddess to Malasana felt clunky. A key element missing was foreshadowing—the gradual introduction of elements needed for Crow pose. Without this, the hip flexion, shoulder protraction, and knee bends required for Crow felt abrupt and unprepared.
“Nowhere does it focus on what it actually takes to lift your feet off the mat and sustain that balancing act,” Schettler pointed out. Essential preparatory poses like Lizard Pose (to press knees into upper arms) and focused gaze cues were absent. Core engagement, crucial for Crow, was also overlooked.
The cues for entering Crow pose were inadequate, and the lack of counterposes (like backbends) left testers feeling off-balance. The timing was also off; the instruction to complete three Sun Salutations in 10 minutes would require holding Downward-Facing Dog for an extended period.
Key Takeaways
- The AI-generated practice felt more like an outline than a complete class.
- AI currently lacks the ability to understand the nuanced connection and flow of a well-sequenced yoga practice.
- The human element of a teacher responding to students’ needs and adapting the practice accordingly is irreplaceable.
- Yoga is a holistic practice that requires a deep understanding of breath, movement, and embodied awareness—qualities that AI currently cannot replicate.
As Bradshaw Wish and Giana Gambino, founders of C.A.Y.A. Yoga School, emphasized, a yoga teacher must trust their body, experience, and intuition—qualities an AI model cannot possess. While ChatGPT might offer inspiration on days when creativity is lacking, it shouldn’t replace the expertise of a qualified yoga instructor.
yoga is the union of breath and movement, a connection that ChatGPT struggles to comprehend. Teaching yoga safely and effectively requires a body that can feel into the shapes and an ability to anchor the practice in something deeper.
Related reading