I Let an AI App Pick Out My Running Shoes—Here’s What Happened

There are few things more personal for runners than finding your go-to pair of shoes. A pair that one person swears by might give another runner major blisters; shoes that make me feel fast and light might make you feel unstable and unsupported.

That’s why trying out running shoes in-person (ideally with the help of an expert) is usually non-negotiable when you’re in the market for a new model.

But what if you could almost replicate that experience in the comfort of your own home?

A new app, Neatsy, wants to do just that, using foot-scan technology and artificial intelligence to make running shoe recommendations, which it then sells in-app. Here’s what happened when I took the app for a spin—and tried the shoes it recommended for me.

The app experience

The Neatsy app is free to download, and taking a scan of my feet took just a few seconds. First, I placed my phone on the ground and hovered my foot above it to take a scan of the bottom of each foot. Then, while sitting down, I scanned the insides of my feet. The results seemed about right: It said that I have a slightly wider than average foot, and very high arches. The app gives two size recommendations: I got 8.5 (which is what I wear in everyday shoes) as the primary recommendation, and 9 (which I typically wear in running shoes) as an option for a more “relaxed” fit.

The app then asked for information about my current go-to running shoes (for me, the Saucony Triumph 19s). Then came the recommendations, which are generated by an algorithm Neatsy created using data from an early version of the app which scanned users’ feet and collected their reviews of running shoes, as well as manufacturer data about shoes (like whether a shoe is specifically made for runners with wide feet), and the preferences I input.

Neatsy recommended 23 shoes from eight different brands, in sizes ranging from 8.5 to 9.5 (since some shoes run slightly large or small). What I was most surprised, and slightly confused by: The wide variety of shoes the app told me to try. It suggested both everyday trainers like the Nike Pegasus 40 and racing shoes like the Saucony Endorphin Pro 3; neutral shoes like the Mizuno Wave Rider 24 and stability shoes like the HOKA Arahi 6; and cushioned shoes like the HOKA Clifton 9 alongside responsive ones like the Brooks Launch 9.

The app didn’t give many details about the shoes, either, or ask me anything about my running experience or goals. (It probably would be helpful for the app to know if someone is looking for an everyday shoe, or something to wear for fast workouts, or a racing flat.) Though I know enough to make sense of the recommendations, I could see how a new runner looking for their first pair could accidentally buy a pair intended for fast marathoners.

Some of the shoes on the recommended list were outdated models, too, a problem that Neatsy founder Artem Semjanow says is due to the fact that the algorithm is based on reviews, and those newer models don’t have many reviews yet. (Usually shoes don’t change all that much model-to-model, but it’s not uncommon for an update to either make or break the shoe for a given runner).

Admittedly, I did not try all 23 shoes that the app recommended for me. I did, however, try a handful, and with the exception of a few strange outliers on the list—which I didn’t try because past experience told me they wouldn’t work—Neatsy’s picks were surprisingly solid. All of the shoes fit well, and while I didn’t have time to put them all through the miles I’d normally like to when trying a new shoe, I logged at least short runs in all of them without any red flags.

Are they all shoes I would have picked had I tried them on at a store? No, and in a way, that may be a good thing. Shoes with lots of cushion have always been my comfort zone, and I’ve been skeptical to try lighter, more responsive shoes because of my high arches and plantar fasciitis struggles. But I was surprised by how much I liked some of the lighter shoes I tried, like the HOKA Mach 5 and the Brooks Launch 9. They may never be my go-tos, but they’ll certainly have a spot in my rotation going forward, and I probably wouldn’t have tried them had the app not suggested them.

What a podiatrist has to say

Can Neatsy replace the in-person shoe buying experience? Miguel Cunha, DPM, a New York City-based podiatrist whom I convinced to try the app, doesn’t think so.

For one, he says, though Neatsy can factor in your foot’s general size and shape, and those measurements can give it clues about your running form and gait, it doesn’t know exactly how you distribute your weight when you walk and run, or how your foot actually lands. Plus, Dr. Cunha says, the app scans the side of the foot while seated—so without full weight on the foot—and thus doesn’t account for the fact that many people’s arches collapse when they are standing. He also points out that it would be easy for someone to mistakenly scan their foot while it’s not in a neutral position—so already slightly supinated (tilting the inner arch up) or pronated (tilting the inner arch down)—which could skew the results.

I also asked Dr. Cunha if he could think of any reason why I, someone with very high arches, would be recommended stability shoes, which are generally designed for flat feet. He said that while there are always exceptions to the rules, and likely plenty of high-arched runners who enjoy running in stability shoes, that’d be something to discover through trial and error, and not something artificial intelligence could predict.

Semjanow says he’s exploring how the app could aid the in-person shoe-buying experience by giving running store employees another tool to help customers find the right shoe, which sounds like a better use of the app than runners using it at home.

But then again, if you’re prone to buying running shoes online sight-unseen, or don’t have access to a local running store, Neatsy could help guide you in the right direction. Artificial intelligence doesn’t know everything—at least about running shoes—but it’s better than nothing.