AI and Running
by Monica Kamin
Artificial intelligence (AI) is becoming part of everyday life, and the running community is no exception. More runners and walkers are using AI to create training plans, answer questions, find running routes, and even stay motivated. But can AI really improve your running? How accurate is it? And where does it fall short?
Over the next several months, the Lynchburg Road Runners Club (LRRC) will explore how AI is shaping the running community. Our goal isn't to argue that AI is good or bad, but to better understand where it can be helpful—and where it cannot.
Before we begin, one important reminder: AI is not a replacement for qualified professionals. It should never be used to diagnose injuries, prescribe treatment, or replace medical advice. Always consult a trusted healthcare professional before making significant changes to your training, nutrition, or recovery.
AI for Training Plans
As we continue exploring AI's impact on the running community, the next topic we want to examine is how to use AI to create training plans. This article is intended to explore how AI generates a training plan—not to evaluate individual execution of said plan. Whether you choose to self-coach or work with a certified coach, remember that training plans are guidelines and may need to be adjusted based on weather, life circumstances, recovery, and how your body is responding. Consistent, smart training is always more important than simply hitting every metric or workout listed on a plan. As a reminder, AI is a tool, not a replacement for professional advice. We highly recommend consulting a trusted healthcare provider, RRCA Certified Coach, or other qualified professionals before making significant changes to your training routine.
Just like in our previous article, we opened an incognito browser window and submitted the same prompts to ChatGPT and Google Gemini to compare their responses. Feel free to repeat this experiment using whichever AI platform you have access to.
Create a marathon training plan for me.
Rather than including the full responses, we focused on the overall approach each platform took. Surprisingly, they handled the same prompt very differently.
ChatGPT generated a generic 16-week plan based on running 4 days a week, assuming an established base of 3–5 continuous miles per run. The schedule included 2 easy run days, 1 speed work session, 1 long run, and a dedicated strength training day. It also provided a list of strength exercises and baseline nutritional advice.
Google Gemini took a very different approach, it advised, “…to make sure it’s safe, effective, and tailored specifically to you, I need a little context first.” It then prompted for current running mileage, timeframe, race goals, and weekly availability. Once provided with those details, Gemini created a 16-week plan and offered a similar mix of workouts, but included specific interval targets, tempo paces, and marathon-pace segments. Gemini’s response was much more thorough and avoided making baseline assumptions about current fitness and goals.
After seeing these contrasting results, we refined the prompt for ChatGPT:
Make me a 16-week marathon training plan that has me running five days a week. My goal is a sub-4-hour marathon, and I am a 29-year-old female.
ChatGPT then generated a much more personalized plan. Both AI platforms peaked with a 20-mile long run, but Gemini scheduled the final 20-miler in Week 12, allowing for a four-week taper, while ChatGPT scheduled it in Week 13 with a three-week taper. ChatGPT's overall weekly mileage was also slightly higher.
How AI Compares to a Certified Coach
Out of curiosity, I compared both AI-generated plans to a generic 20-week marathon training plan created by an RRCA Certified Coach. The coach's plan called for five running days per week and fell between the two AI plans in overall weekly mileage. It also includes 20-mile long runs, with the final one occurring three weeks before race day.
The biggest difference wasn't the mileage—it was the level of detail. The coach's plan included structured post-marathon recovery, more specific workout guidance, and two one-mile time trials to measure progress throughout the training cycle. While the overall training pattern was similar, the coach's plan offered greater depth and progression.
Testing a Different Distance
Next, we tested a different scenario by opening a new incognito browser window a few days later to see how AI would create a 5K training plan. We intentionally waited a few days and used a new incognito session to ensure the results weren't influenced by previous conversations or chat history.
I am getting back into running after taking a year off. Create a 5K training plan for me.
Again, the platforms responded differently.
ChatGPT created an eight-week program with four running days per week that relied entirely on mileage-based workouts. Several runs reached 5–6 miles—nearly twice the race distance.
Gemini created a six-week program with three running days per week and primarily used time-based workouts, such as a 20-minute run/walk instead of prescribing a set distance. Only a few workouts were distance-based, and the longest run was 3.25 miles. Unlike when we asked Gemini to create a marathon training plan, it did not request additional information before creating the 5k plan.
So, What Does This Mean?
Even among experienced coaches, there is no single "correct" training philosophy. The Hansons Marathon Method emphasizes higher weekly mileage over six days of running, while Jeff Galloway's method centers on the run-walk approach with gradual long-run progression and structured recovery.
The AI-generated plans were not necessarily bad. For healthy runners looking for a free starting point, they can be a useful resource. However, they may not be the best option if you're new to running, training for a new distance, or returning from an injury.
Experienced runners may get the most value from AI by providing detailed prompts that include current fitness, weekly mileage, race goals, schedule, and preferred training style. Even then, working with a trusted coach or qualified professional may provide a better path toward ambitious goals, such as qualifying for the Boston Marathon or breaking 18 minutes in the 5K. There is also something irreplaceable about working face to face with a trusted coach and that experience cannot be replaced by the training recommendations AI provides.
If you choose to use AI to build a training plan, remember to be as specific as possible with your prompts, pay attention to how your body responds, and don't be afraid to adjust. If you prefer time-based workouts over mileage, make that clear. If five running days becomes too much, scale back and ask AI to adjust the plan based on four days of running. Don’t forget to include appropriate cross-training and recovery so the plan fits you.
Next month, we'll explore how AI can be used to evaluate running gear recommendations. In the meantime, if you'd like to learn more about AI or explore different training methods, our board members have a list of recommended books we'd be happy to share—just ask!
