A shoe can make or break your run, no matter your fitness level. Ill-fitting shoes can cause blisters so bad you can’t run anymore or rub your toenails right off. Or the shoe just doesn’t serve your specific gait, compromising your performance.

That was the case when pro marathoner Nell Rojas made headlines for breaking a coveted contract with Adidas. She cited issues with Adidas’ Adizero Adios Pro 2 for the split in 2022. 

“I know that shoe is an amazing shoe, and I know world records have been set in it and marathon winners run in it, but for whatever reason it didn’t work for me and my stride,” she told Women’s Running. “I raced in it and didn’t have good results. I was having a hard time recovering and dug myself into a huge hole, and I think a lot of it was because of the shoe change earlier this year.”

What if runners could save themselves from training in shoes that are ultimately bad for them? What if a software, powered by artificial intelligence (AI), could recommend the perfect shoe for you?

Why Is the Perfect Shoe so Elusive?

While pros are bound to a single brand, the rest of us are burdened with choice. Do you choose whatever your speediest friend is wearing? What looks the best? Do you choose one that has excellent online reviews? Or visit a running shoe store and rely on a random salesperson’s expertise? 

Maybe you’ve done all of the above and still felt disappointed with the performance of your footwear. For many runners, finding the right shoe is tricky and frustrating. It’s also paramount to success in the sport. 

“It’s the only equipment for running that needs to be perfect,” says Tim McConnell, who owns the store, West Seattle Runner alongside his wife Lori. Together, they’ve been fitting runners in shoes for more than a decade. Lori is also a certified running coach and licensed mental health counselor. The wrong shoes can not only make running uncomfortable, but can cause serious injuries up the chain. 

So why is the perfect shoe so elusive? For one, just getting sizing correct is a complicated mess.To start, our feet change over time. As we age, feet flatten as ligaments and connective tissue loosen, causing feet to change size. Hormones during pregnancy do the same.

“Even as you get into that run, your foot swells,” says Mark Bouma, a Seattle-based physical therapist. Which means that your feet will be smaller in the store when you’re trying on shoes than they will be out on the pavement. 

Sizes are also not precise and vary from brand-to-brand. In women’s running shoes, for example, it is estimated that 40 percent are not true to label size according to data recorded from Volumental–a company that makes 3D foot scanners for shoe stores. 

Enter AI 

Many running stores are equipped with 3D foot scanners that measure heel-to-toe-length, ball width, ball girth, heel width, arch height, and instep height. The scanners help store employees recommend shoe size, can generate custom insoles based on the readings, and use AI to recommend shoe models. 

While this is a quick, simple, and free service that stores provide to the community, it’s an extra step that some consumers aren’t interested in or otherwise don’t have access to if they don’t live near a retailer. Instead, consumers purchase shoes online, hope they fit, and return if they don’t. It’s a cycle that causes friction for all parties–retailers, manufacturers, and consumers. 

No one is surprised to hear that the trend of online shopping is not likely to go away. As Forbes reports, eCommerce continues to grow steadily and is expected to balloon to 24 percent of all retail sales by 2026. 

What if there were a way to buy shoes online better? According to research from Salesforce, product recommendations powered by artificial intelligence can potentially affect $194 billion in online sales. 

Artem Semjenow, founder and CEO of the running shoe app Neatsy.ai agrees that artificial intelligence and augmented reality can ease pain points for shoppers. His app allows users to scan their foot from their phone to receive the kind of recommendations they’d get in a running store as well as physical therapy recommendations related to any pain they are having. 

As someone who suffered from debilitating foot pain, Semjanow, wanted to create a tool that would help people, not just runners, get access to affordable foot care and personalized shoe recommendations. Prior to founding his tech company, Semjanow was at Prisma Labs developing AI photo-editing software.

“I decided that I got tired of improving people’s selfies. It’s a cool job, don’t get me wrong. And people pay for self-improvement. But I wanted to make something more meaningful for the world.” 

Through his own experience and what he knew from doctors in his own family is that when you go in with a problem, doctors will always try to take a more conservative approach to addressing your problem before suggesting an invasive treatment. What if he could develop an app that went through those conservative approaches to save users time and money? What that looks like on Neatsy.ai is giving guidance about basic foot ailments, suggesting physical therapy treatments, and recommending the shoes that you can purchase from the app. 

The system uses augmented reality to create a 3D scan of the foot and artificial intelligence to make the recommendations and answer questions users have, similar to ChatGPT. Arguably more reliable than ChatGPT, Neatsy.ai has been trained using real medical cases in partnership with Harvard Foot and Ankle Research and Innovation Laboratory. 

“They helped us collect the anonymous set of data of different cases for people—pretty much all the 3D foot shapes, and what real doctors have thought that the problem might be and what treatment and shoes doctor recommends in that particular case,” says Semjanow. 

The team will soon publish research on the process of training AI using a mix of in-app scans and obtained medical cases. 

What Is AI Missing? 

When a store associate is conducting a shoe fitting, they are taking a lot of qualitative data into consideration: your running goals, your history with injury. They are watching you run in the shoe to see how your gait is affected. 

“We even listen to the sounds of the runner,” says Lori McConnell. If the footfall is particularly loud or slapping, that might lead them to recommend a different shoe.

Shoe fit is an intricate puzzle when you consider the dynamic nature of the foot, says Craig Fox, D.P.M., an Illinois-based podiatrist. Then you add in the extra layer of performance goals or other tangible variables. “We know the size of your foot changes dynamically as you walk or run. What if you change socks? That may have an effect on shoe size. Even with 13 points of measurement, the challenge persists.”

It’s unlikely that any scan will understand how your foot moves under load, unless it is taken in motion, says Bouma. “If it’s just scanning your foot, you really don’t know what the other joints around the foot are doing.” 

Another complication comes in the changing landscape of shoe design itself. Jon Teipen, principal footwear product line manager at Brooks, notes that shoes are more dynamic than they once were, which changes what we know about fit. 

“You’re looking at more engineered materials that are a lot more intelligent, so they just have a little bit more give to them,” he says. In older designs, shoes would have a repeating mesh pattern with polyurethane (PU) overlays that made the upper more static and less accommodating to different feet. “[Now,] as you put your foot in there, the materials can expand out a little bit more than they used to be able to with all the extra overlays on it.”

He believes that extra level accommodation gives AI an advantage in spitting out a closer-fitting recommendation, as long as the internal measurements are right. 

“On the other side, too, is that the materials have some flexibility and stretch to them. There’s some variants of the actual shoe structure itself that the program’s probably not going to be able to decipher either,” says Teipen. Dr. Fox notes that the same issues persist with foot scanners in store that are designed to narrow in on foot size and not much more. “Just because the shoe ‘fits’ on your foot, does not mean it is the best footbed for your foot type.” 

‘A Footscan Gets You Halfway There’

Even running store owners agree in regard to the tech they use everyday. 

“The scan gives us credibility,” says Tim McConnell who is used to customers not believing him when he suggests they’re wearing the wrong sized running shoe. But many of his recommendations are based on his experience and not necessarily because of the technology. “A footscan might get you halfway there,” he says. 

A lot of that push back has to do with lack of education–adults not understanding that their feet can get bigger–and what industry experts call ‘vanity sizing’ –consumers holding on to a standard of femininity or masculinity attached to confining within a certain shoe size. 

Teipen also sees how a technology like AI can help runners get beyond vanity sizing issues. “Tools like this that are a little bit more objective, might be good for people to get over any mental barriers they have.” 

Semjanow also recognizes the limitations of Neatsy’s use of AI in such a broad category. With so many new shoe models released each year, the app is unable at the moment to recommend the newest, perhaps most advanced shoe models, as it relies on user reviews to generate recommendations. 

“For example, HOKA introduced a new running model for summer 2024, and we just don’t have any data for that,” says Semjanow. Which means app users are more likely to be given recommendations for older models, rather than new ones. 

“We want to establish more partnerships with shoe brands to actually get the data for different kinds of shoes before they are launched.” 

It’s important to note that the company does receive a commission from shoes sold on the platform. But according to Semjanow, the commission is the same across the board, so there is no incentive for them to push one brand over another. 

An AI Wishlist 

AI may not be the perfect shoe conversion tool right now. But there are other ways AI will revolutionize the running shoe industry. 

“The future of footwear lies in reimagining the selection process, focusing not on the shoe as a whole but as a sum of its parts,” says Dr. Fox. “Analogous to selecting eyeglasses, where frames are chosen and lenses are made to specifications, the concept of prioritizing the insole before the shoe itself could revolutionize the shoe buying experience.”  

Brooks is in the early stages of utilizing AI in a generative design process. 

“It’s starting to open up where we can move faster,” says Taipen. When working on a blueprint of a shoe, for example, shifting from a 2D to 3D model, these technologies allow them to modify characteristics faster than the modifying hand drawn designs. “The hope is, as we move into the future, to be able to do actual digital testing.”

With the forthcoming Hyperion Elite 4, for example, Brooks has partnered with Arris, a manufacturing tech company focused on performance products, for the shoe’s carbon plating. 

“What they’re able to do is lay down each fiber individually, so they can add a little bit more stiffness in some spots, a little bit more flexibility in some spots. And then they can run it through their computer and test it and see what the performance and functionality would be. So they can do a lot of iterative testing before actually even building a part.” 

Teipen is encouraged that generative design in AI will only get better.Likewise, Dr. Fox is hopeful about AI and AR in shoe manufacturing and retail. 

“After three decades in podiatry, I’ve witnessed countless patients lugging shopping bags filled with shoes that didn’t quite work. Their stories often echoed the frustration of finding a temporary solution, only to return to the online marketplace in search of the elusive perfect shoe–the unicorn of footwear.”