\nSpecificationDetailsPrice$9.99\/month (premium tier upgrade)Core FeaturesAdaptive machine-learning training plans and high-definition video routesCoaching ArchitecturePredictive pacing and fatigue modeling developed by Ryan Cooper (Best Bike Split)Algorithm Data InputsPeer-reviewed sports science combined with anonymized power and heart rate data from millions of usersVideo Routes at Launch21 routes across 6 countriesTotal Video Distance385km of new roadsVideo GamificationRetains PowerUps, Segment Leaderboards, HoloReplays, XP, and Drops[cite: 20]Software RequirementsZwift Companion app (required for plan setup, progress tracking, and daily adaptations)For More InfoZwift Website<\/a>
\n
![\"Zwift](\"https:\/\/velo-cdn.outsideonline.com\/wp-content\/uploads\/2026\/10\/DSC_7382.jpg\")
<\/a>(Photo: Josh Ross\/Velo)<\/p>\nToday, Zwift is launching Zwift Plus. This new premium tier pushes the monthly cost to $29.99, with a discounted yearly option at $299.99. In exchange for the upcharge, you unlock two core upgrades: real-world video routes and adaptive coaching.<\/p>\n
Zwift considers this a massive milestone, which is why I recently went to Moab to celebrate the video-based routes ahead of launch. The concept was to ride the route indoors, then ride it outdoors to see how perfectly they matched. Unfortunately, we caught the tail end of a Pacific hurricane, so I only got to drive the route and ride it inside. Still, the big reveal ended up feeling almost anti-climactic\u2014but in the best way possible.<\/p>\n
The inside perfectly matches the real world, but you still get all the Zwift details you are used to. Zwift lays the graphics seamlessly into high-resolution and high frame rate footage. I rode the Tron bike in Arches National Park, and while it sadly doesn\u2019t glow on the road, the wheels sit perfectly on the video road and the speed adjusts as you ride without ever breaking the illusion. It\u2019s underwhelming, but only because it\u2019s really good and feels like a natural extension of what Zwift already does. If you\u2019ve used Rouvy, which Zwift now owns, this is better.<\/p>\n Personally, though, I don\u2019t find the video-based routes all that much of a draw. They provide more variety, and that\u2019s great, but Zwift already adds routes all the time and anyone who\u2019s used Zwift knows the team is great at recreating real-life locations. What this launch is really all about is coaching, and that\u2019s where the real magic is.<\/p>\n Part of why the team at Zwift considered this launch such a high-stakes proposition is that it\u2019s a price increase, and that always comes with baggage. You could even consider the addition of the video routes to the Zwift Plus plan as a sort of tiered pricing structure where, all of a sudden, you have to pay for more routes. That\u2019s not really true, but I can understand the hesitation.<\/p>\n I think part of why I don\u2019t see that as an issue is because I don\u2019t find the video routes to be the major news here. They are additional content, and as someone who will likely chew through every single additional badge Zwift offers this winter, that\u2019s nice, but also not a big deal. Instead, I see this as a pretty clear value proposition of $10 a month for structured training.<\/p>\n With that in mind, my favorite thing about Zwift Plus is that I see it as a product that is going to consolidate the structured training market. I believe this will be a big hit to TrainerRoad, and I think that for many people, Zwift Plus will represent a net decrease in monthly subscriptions.<\/p>\n The reasons for this are pretty simple: structured training has to be good enough, and it has to create an environment where the user actually follows the plan. Before this launch, Zwift was mostly failing at both of those requirements.<\/p>\n Zwift already had training plans; they just weren\u2019t engaging, they weren\u2019t reactive, and they were widely considered to be not that great. With Zwift Plus, the first correction was the quality of the plans, and to address that, Zwift brought on Ryan Cooper.<\/p>\n If you aren\u2019t familiar with the name, Ryan Cooper is the founder of Best Bike Split and the former Chief Scientist at TrainingPeaks. To understand why his involvement is so important, you have to look at what Best Bike Split actually does. It is a predictive physics engine that takes a rider\u2019s specific power profile, weight, and aerodynamics, layers it over the exact elevation of a race course, and calculates exactly how many watts that rider should push on every single segment to finish as fast as possible without blowing up.<\/p>\n That is the level of data-driven modeling he brings to Zwift, and it is exactly why the brand is very careful to call this new coaching system \u201cmachine learning\u201d rather than \u201cAI.\u201d Right now, AI is a marketing buzzword. When most people hear it, they picture a chatbot scraping the internet and hallucinating a generic training plan based on a text prompt.<\/p>\n Machine learning is different. It relies on math and datasets. In this case, Cooper\u2019s algorithms were trained on the anonymized power and heart rate data of millions of Zwift riders, combined with peer-reviewed sports science. The system isn\u2019t guessing or estimating based on what it read online. It uses actual machine learning to recognize patterns in your specific history, power data, and daily fatigue, and then mathematically calculates exactly what you need to do today to actually get faster.<\/p>\n This means Zwift has now solved the quality problem that TrainerRoad has so far always held as an advantage. But the raw quality of workouts is actually somewhat secondary in real life. You need not only quality workouts, but you need workouts that people will complete, and this is where Zwift has something no one else does.<\/p>\n Zwift controls everything about where you ride your bike, and that means the Zwift Plus workout plan has options. Once you set up your plan, the system will recommend an ever-changing variety in terms of how you complete a workout. My favorite example was actually the first day of the 12-week plan I started, when I was scheduled for a route rather than a set of intervals.<\/p>\n Except, if you squinted a bit, you\u2019d see a clear interval workout. The route was the Watopia Mountain Route, and while it might appear to be a free ride, the route starts with a few miles of flats, then it climbs for a while before descending and finishing with some flats. What is actually happening there is that you get about 8-10 minutes of warm-up, then you have a work block of about 40 minutes before you get a rest, and finally a 20-ish minute cool-down where you ride the flats again. And Zwift knows, based on all your past rides both inside and out, exactly how hard you will ride each of those sections. Except I actually went harder because I started late and wanted to finish faster, so in my case, I got a notice asking if I wanted to adjust my plan based on the new intensity.<\/p>\n In the first week, I also had traditional intervals and a follow-the-RoboPacer session. The result is a tight training plan that adapts to you if you don\u2019t follow it exactly, but rarely looks like a tight training plan. It\u2019s a sort of trick to keep you engaged, and my money is on it working very well.<\/p>\n The only problem with this system is that, at launch, there\u2019s not enough AI and there\u2019s not enough feedback. I am told some of this will come, but it\u2019s not super clear when or how much, so I have to evaluate what exists right now\u2014and right now, it needs more interaction.<\/p>\n I am sure many people will not want to hear that this should have a chatbot, but I think it should. It has to be a good one, but for more than a year now I\u2019ve bounced around between different AI systems, and when it works, it\u2019s incredibly helpful. For a long time, I used Google Gemini paired with Garmin Connect, and this summer I\u2019ve been playing with RideWithDave. Whatever system I\u2019ve used, the main benefit is that you can ask the robot about deviations. \u201cI have this scheduled; what if I do this instead? Does that fulfill the same goal?\u201d Or, the opposite happens plenty, too: \u201cHey, I had this scheduled but instead I did this; did I satisfy the goal?\u201d With the Zwift system, you just have to guess.<\/p>\n And once you do guess, there\u2019s very little feedback one way or the other. One day, I had an interval session scheduled that was 2\u00d720 over-unders, but I opted to go outside instead. Every chance to ride outside in the fall is borrowed time (it\u2019s not pouring in Portland yet), so I went out and did a 1\u00d740 interval at roughly the same power. Because I talked to Zwift about it directly, I know I achieved very close to the scheduled workout. However, the Companion app doesn\u2019t give you that feedback, and it refuses to give you a green checkmark for the day despite the fact that you completed the physiological goal of the workout.<\/p>\n That same opacity exists when the system recalculates your plan. Zwift was smart not to create a system that automatically recalculates in the background, which would essentially make it feel like you don\u2019t have a plan at all. Instead, when I went harder than expected on my first day, it explicitly asked me if I wanted to adapt the plan. I clicked yes, but it wasn\u2019t clear what actually changed.<\/p>\n The next time it happened, I had a 2.5-hour ride planned at the top of zone 2. Instead of doing it as a single workout block, I broke it up across three routes using the new Wattlock system (an ERG mode that is now available during Zwift free rides). I completed the required work, but again, I didn\u2019t get the green checkmark. Then, because I did an extra 30 minutes of easy riding while chatting with a friend, the app prompted me again asking if I wanted to adapt the plan. Thirty minutes of easy spinning doesn\u2019t seem worth overhauling a schedule for, so I haven\u2019t responded to the prompt, but it leaves you second-guessing. If this system is so smart, why would it ask me to adapt the plan if I didn\u2019t actually need to? A conversational chatbot could explain this immediately, but barring that, it could also be fixed with some simple contextual cues\u2014and by just giving me my green checkmark.<\/p>\n Some of what I don\u2019t like is probably livability too but it goes a little farther. Zwift knows your FTP but it will only raise it, not lower it. In theory that means I have to do FTP tests despite Zwift already knowing what it is. I can\u2019t say for sure that\u2019s necessary but I did one before I got started and that\u2019s an easy fix. I also want more info within Zwift for outdoor rides. Right now it just links to Garmin or Wahoo, and it does have the info in the backend, but I want my outdoor rides to give the same ride tracking screen as my indoor rides. Just build it all in.<\/p>\n Other than that, you\u2019ll find a new look to the Zwift Companion app and the plan lives in there. It\u2019s all quite straightforward for setup and if you need to schedule days off for any reason you can do it. You\u2019ll get a prompt to adapt the plan, as I\u2019ve already discussed, and you can decide if the changes mean you need a change to the plan. Everything here works really well.<\/p>\n<\/a>(Photo: Josh Ross\/Velo)<\/p>\n<\/a>(Photo: Josh Ross\/Velo)
\nWhat I like about Zwift Plus<\/p>\n<\/a>(Photo: Josh Ross\/Velo)<\/p>\n<\/a>(Photo: Josh Ross\/Velo)
\nWhat I don\u2019t like about Zwift Plus<\/p>\n<\/a>The Watopia Mountain Route is analogous to an interval workout it just doesn\u2019t look like it. (Photo: Josh Ross\/Velo)
\nLivability<\/p>\n