Peloton is using AI to keep you from canceling

My old apartment had a gym in the basement with a Peloton machine. I was intrigued. I wanted to test it out and see what the hype was, but you needed those clip-in shoes to use it. So, and I’m embarrassed to admit this, I bought the shoes just to try the machine.

My first Peloton ride was a lot of fun! I didn’t realize how hard a spin class would be. My quads were on fire, but the music was bumping, and I had the virtual instructor to motivate me and push me.

I did roughly three more classes before I forgot all about it. I still have the shoes, years later, somewhere in the back of my closet, because maybe, just maybe, I’ll try it again one day.

It’s hard to maintain a consistent fitness habit. It’s even harder when you don’t have a set plan and are just choosing from a catalog of random workouts. This week, I’m breaking down how Peloton introduced an AI layer called Peloton IQ that has actually increased user engagement and retention.

TL;DR

The thirty-second version of what Peloton did.

Peloton sells bikes and treadmills that cost about as much as a used car, but the real business is the $44-a-month membership. The bikes are really the on-ramp to that subscription, which means the whole game is keeping you on the machine after your New Year's motivation wears off.

In October 2025, Peloton launched Peloton IQ, an AI coach that works before, during, and after your workout. It runs on any Peloton, as well as on the app. It builds a weekly plan around a goal you choose (get stronger, boost your cardio, or simply live longer), predicts how challenging a class will be for you, and sets your target numbers before you start, then reads your results afterward and recommends the next workout. If you get one of their brand-new plus models, which has a movement-tracking camera, it counts every rep so you don't have to, calls out your form in real time ("keep your back straight"), and tells you when to grab a heavier weight. Free brain for everyone, premium eyes for buyers.

Almost half of all users tried Peloton IQ, and after a price increase pushed monthly churn up to 1.9%, it fell to 1.2% the next quarter, with the CEO crediting the new AI layer. The most-used feature Peloton has shipped in years is also the reason fewer people are canceling. The base is still down about 7% YOY, though, so this is Peloton keeping the members it has, not yet winning new ones.

Under The Hood

The three strategies that make it work.

1. Personalize your user’s experience

Peloton shipped IQ's personalization layer, the plans, the weekly insights, and the "here's what to do next," free to every machine it ever sold and to app members, too. Most Peloton bikes, which have spent years as a very expensive coat rack, suddenly had opinions about their users’ cardio (or lack thereof).

The point of giving it away is retention. Peloton makes its money on the monthly membership, so a member who feels coached is a member who keeps paying. And the engagement showed up fast: 46% of active members used IQ's insights in its first quarter, more than 600,000 started a Personalized Plan, and engagement with those plans climbed more than 10% in a single quarter. Engagement was rising more broadly, too, with workout time per member up 7% from last year. Peloton had raised prices in the fall, which pushed monthly churn up to 1.9%, then down to 1.2% the next quarter, with the CEO crediting "personalized plans from Peloton IQ."

Why did it drop? I think it’s because creating a tailored plan and giving recommendations on future workouts reduces decision fatigue. When AI is properly embedded into the product experience, it can reduce the work that a user needs to do (we noted the same thing in our Spotify article). It’s easier to stick to a plan when you have a plan, and the AI does the heavy lifting to provide it.

2. Create a data flywheel

IQ improves with data, and Peloton feeds it two kinds. Your own: every ride, run, and lift you do on the equipment, plus anything you connect from Apple Health, Garmin, or Fitbit. And everybody else's: the workouts of millions of members who train something like you.

If you don't wear a tracker, IQ still personalizes for you. It estimates your effort by looking at members who wear one and train like you, then works out numbers like your cardio fitness score from there. So somewhere out there, a stranger with a Garmin and more discipline than me is doing sessions that teach my bike what good effort looks like. IQ can coach me on the days I tell it almost nothing, because it's leaning on everyone who came before me.

That's the data flywheel. More members means more data. More data makes the coaching more personal. Better coaching keeps people engaged and paying. Higher retention funds more members and more data. The plan you get a year in fits better than the one you got on day one.

Every member makes the coach smarter for the next one.

3. Collect more data, collect more $$$

The most advanced parts of IQ, the camera that watches your squat, counts your reps, and tells you to grab a heavier dumbbell, only work on the new plus models. That movement-tracking camera is why Peloton can charge more for them. The base Cross Training Bike is $1,695, the Bike+ with the camera is $2,695. The extra thousand dollars gets you the IQ that can actually watch you work out, and yes, judge your squat depth. That's what people are paying for.

You don't need hardware to run this play. I wrote about Duolingo doing the software version: its best AI, the GPT-4 Video Call and Roleplay, lives in the pricey Max tier, while the simpler stuff stays free. Peloton's twist is that its gate is a physical capability, a camera that the cheaper bike doesn't have, so the upgrade reads as a real reason to buy instead of a fee to unlock something the old machine could already do.

The biggest takeaway here is that the more data you can collect on the user, the more helpful your AI can be, and possibly the more you can charge. For a hardware company, it might mean adding more sensors to a device, and for a software company, it could mean asking the user for more permissions, like location or health data.

That’s all I got for this week!

Two asks before you go:

1. Reply to this email with any product features that have kept you from leaving (for me, it’s the Duolingo Owl threatening me).

2. I’m trying to grow this newsletter, so please share it with your friends, coworkers, and frenemies!

— Amaraj (aka catch me on the Peloton leaderboards)

The Meme