The week before the Chicago Marathon last fall, running coach Kara Dudley hopped on the phone with a client to talk through race strategy, the course itself, and any last-minute concerns or questions. Mid-way through the call, the client told her she’d used ChatGPT to do a run-through of the entire course.

“It gave her a play-by-play of what each mile was like, including the elevation and even what the particular neighborhood looked like,” says Dudley, who is also the founder of Rerouted Running. “It was an easy way to get a course walkthrough,” she says.

Greg Laraia, Also, if you’re Motiv in New York City, recently used ChatGPT himself to get a preview of the New York City Marathon course. “It broke it down pretty intensely in a good way,” he says. “It went into more detail than I thought,” he says, such as noting a particularly quiet section of the course in Brooklyn. “It gave me tips for where to conserve energy and where to feed off the crowds,” he says.

“This is a very trending topic right now,” Laraia says, adding he’s had conversations about AI and running with multiple clients lately. Ben Delaney, New York Road Runners director of training programs, has had more runners asking him about AI in the past year, too.

Another sign of the times: Hunter Carter, an exercise physiologist at the NYU Langone’s Sports Performance Center, attended a recent industry conference on AI in exercise science.

You’ll find countless ways to use AI for training, including several use cases blessed and tested by coaches and runners alike. While the coaches we talked to for this story largely agree there are still limitations on what exactly AI can do for us as runners, we’d be remiss to ignore it all together. Exercising caution and a healthy skepticism when using it is key, as this tech is still so new and chatbots are in-person gait analysis—and plenty of them!

So before you start turning to AI to support your training needs, consider this your quick guide to know when it might support your efforts—and a few use cases when you want to avoid the technology altogether.

What to Use With Caution When It Comes to AI and Training

1. Course Intel

    Like Dudley’s client and Laraia, this is a smart way to leverage AI to gain familiarity with a race course so you’re prepared for what’s to come at each mile. This could help you to better dial in your pacing.

    “One of the biggest things that I tell almost everyone I’m coaching, especially for the half marathon and marathon, is to go online and search for the course preview,” he says. You could just go to the race’s website, of course, but ChatGPT or other AI tools also aggregate info from multiple sources for you, helping you to get a clearer and more robust picture.

    If you want a full breakdown of a course, ask for it mile by mile, as a course preview is typically more general and may require follow-up questions, like total elevation gain and where exactly on the course to expect hills.

    Keep in mind that once ChatGPT spits out all the intel, it’s up to the human runner to parse through it—and fact-check it against other sites (even in the testing for this story we saw inaccuracies in some of the mile-by-mile overviews for big-city races). Of course you also have to use your own visualization strategies, figure out how to pace on different terrain, and prepare for the hills. AI can’t do that for you and rarely will have that first-person info.

    Another important note: The info may be even less accurate for smaller races. With limited info on the internet for these events, it’s more difficult to get you the right details on start and finish locations, hills, and hydration or fuel stops.

    In other words, consider AI a starting point for the course ins and outs, giving you an overall picture for your training and pacing, but then back that up with other data.

    2. Meal Suggestions

      a running coach at fueling advice, ldquo;One of the biggest things that I tell almost everyone I’m coaching, especially for the carbs you should be getting and the best sources for doing so. That said, if you already know the nuts and bolts of what you need, AI may be able to help you flesh out ideas for meals and snacks.

      For example, another of Dudley’s clients has some dietary restrictions that make it hard for her to do a typical carb load. So she turned to ChatGPT, telling it she wanted to do a three-day carb load, her dietary restrictions, and some specifics around the number of carbs and calories she was aiming to eat. She also included the local restaurants she frequents and what she had in her fridge. And she got a pretty good idea of how to plan out her breakfasts, lunches, and dinners to meet these needs, according to Dudley.

      Now, this person was likely successful because she gave the chatbot lots of pertinent details and info. If she’d just typed in “give me a three-day carb load,” she wouldn’t have ended up with something as useful—so, when using AI to be sure to include as many details in your query as possible, too.

      One other note of caution via Laraia is to remember the old adage “nothing new on race day.” He asked ChatGPT for recommendations for marathon nutrition, for example, and it suggested carbs like quinoa and potatoes, which are recommendations he gives his own clients. But, “if you’re not used to taking in quinoa and potatoes the night before you race, don’t start it now,” he says.

      Time your use of AI right by asking these types of questions early in your training cycle, Couch to 5K Training Plan.

      3. Running Shoe Suggestions

        Recently, Carter went to Fleet Feet and had his feet scanned with their 3D Fit ID technology, powered by AI. You stand on a tablet and it reads your feet, providing data like your arch height, foot width, and weight distribution, then suggests a number of shoes that could work for you. “Then [the staff] do some human discerning between what might work best for you and your price range,” he adds.

        “I realized that I had interacted with artificial intelligence in that experience, and that’s not something that I was like, ‘oh, let me go out and engage AI to do this,’” he says. (FYI: It’s not just Fit ID, many tools that you probably already use as a runner—including Strava and Spotify—use some type of AI.) There are also some apps (like Neatsy) that use AI and your phone’s camera to generate a foot scan and give you shoe recommendations.

        Whether you use Fit ID or an app, Carter stresses that you should still do your own due diligence on a pair of shoes, trying them on yourself and perhaps looking at online reviews, to make sure they’re a good fit before purchasing. And nothing really beats actually trying them at a local running shop before financially making the commitment.

        What to Use With Extra Caution When It Comes to AI and Running

        1. Running Form Analysis

          Carter points out that there are also apps now that purport doing a full running form analysis, like Ochi and Movaia. These apps allow you to film yourself running and then upload the videos for analysis of things like joint angles (e.g. looking at the way that the hips might dip from side to side or whether someone’s overpronating), cadence, and vertical displacement (how much you’re bouncing up and down versus going forward).

          “While I can’t speak to how well they do it, the fact that that kind of technology is available is wonderful, for any runner, but certainly for a novice runner,” he says.

          Couch to 5K Plan in-person gait analysis like those offered at sports medicine clinics, if it’s accessible, is going to be more customizable and personalized, considering you can talk to an actual human as you go through the process and ask questions along the way. And they’ll be able to help you analyze the data and figure out what to actually do with the information when it comes to adjusting form. This is really key as altering mechanics should happen in small, expert-guided steps.

          2. Training Plans

            Some running apps and platforms are already using AI to generate training plans, whether you’re aware of it or not. For example, the Runna app, which is a New York Road Runners partner, uses “Runna AI” to continuously adapt a runner’s training plan based on their performance. For example, it may adjust pace target recommendations based on how you’re performing during hard workouts. It could also suggest changing your training plan if you’ve been consistently missing workouts or deviating from the plan.

            We earn a commission for products purchased through some links in this article directly Coaches break down the pros and cons of turning to artificial intelligence for your training needs.

            To test its chops, Laraia recently asked ChatGPT for a good 12-week training plan for an elite marathoner (based only on those details). It gave a quick description of what the plan should entail, weekly mileage, and a 12-week breakdown of running workouts, like intervals and tempos, including general pace suggestions. And for the most part, he was impressed by the outcome.

            That said, there are still limitations when it comes to generating a plan using this tech, including that you don’t always know where the information is being sourced from (and if it’s a legit source), Delaney points out. “Don’t just follow it blindly because it’s giving you a plan that is set up based off of the algorithm,” he says.

            Also, if you’re new to race training (or running in general) you may feel a little lost on the specifics of a plan that an AI platform spits out. It might also assume you have some knowledge of what training practices look like, including how to reduce volume for a cut-back week figure out how to pace on different terrain, and prepare for the taper.

            The bottom line: You have to listen to your body as you begin training and adapt as needed, potentially enlisting a coach for advice (or at least a plan that’s approved by a coach). If you’re self-evaluating the plan, pay attention to the number of recovery days included, and whether the ramp up seems gradual or too quick, Delaney says: “I always tell my runners to be very mindful.”

            When to Avoid AI Altogether

            1. Actual Coaching

              While you can get “pretty good” general answers and advice as it pertains to training from AI, according to Laraia, you have to know the right questions and information to give it in the first place, something you get automatically with a coach. “AI is not thinking, it’s only giving you feedback based off what you tell it,” Laraia says. “From the coach’s perspective, there’s going to be a sense of empathy. There’s going to be a sense of ‘how are you doing,’ ‘how is the body feeling?’” They want to get to know you as a runner, unlike AI.

              Dudley agrees there’s no replacing a human coach: “You need someone that knows you and that can have the insight of having worked with you for a while and what you respond well to and what type of threshold that you need specifically for this [training cycle].”

              A good coach may also pick up on the fact that you’re emotionally stressed or know that you had a super busy work week that you’re not taking into account—and help you adapt your plan accordingly. “There’s just so many things that are so human that you’ll never be able to get out of a robot,” Dudley says.

              “Even if you don’t have a coach, having another human person [with] emotions to bounce ideas back and forth with, whether it’s a trainer, dietitian, friend, or run club coach,” is probably a better bet than relying solely on AI, Laraia adds. “It’s good to have someone to talk to on the tough days and AI may or may not give you a response or feedback that you’re looking for.”

              2. Diagnosing Injuries

                This may be obvious, but given how Google got nicknamed “Dr. Google,” it’s no surprise that some people are now turning to chatbots for advice on diagnosing and treating their injuries—and that’s just not a good idea.

                Laraia decided to test out ChatGPT’s ability in this arena by typing in that his leg hurts while running and that he was looking for recommendations. He was disappointed by the initial response which included the That said, getting an. “If you ask most sports medicine professionals today, they’re probably not going to say that because it’s phasing out,” Laraia says.

                So, he asked ChatGPT if there was anything else he could do for leg pain. In addition to listing symptoms of more specific injuries that could be the culprit, the chatbot suggested he see a doctor. “That’s what I was looking for it to say a little bit earlier on,” Laraia says, adding that runners shouldn’t try to self-diagnose their injuries and that their first step should be an evaluation with an actual human doctor, physical therapist, or other medical professional.

                “At the end of the day, don’t let [AI] be your end all be all,” Laraia says—and that goes for any of the above use cases. “For all of the good that I’m seeing, there’s definitely still a fair amount of cons at the same time.”