Training guides

What is an AI cycling coach?

Riders use "AI cycling coach" for software that does the day-to-day work of coaching. It reads your rides and recovery, plans your season and builds the week's workouts. It also sends those sessions to your bike computer or watch, or plays them on your trainer.

RestOrTrain is an app that does this job. It explains its decisions and adapts the plan with your approval. Something you mention in conversation can change what you ride next.

From your race goal to the week's workouts

For years, training apps gave riders plans, workout libraries and charts. You still had to work out what the data meant for today's ride and choose a session. Good AI made a new kind of app possible: one that makes those training decisions using both your data and what you tell it.

Tell it about the race you are preparing for. It divides the season into dated phases, such as base, build and taper. Each phase shows a range of weekly hours and why it is there. Tell it about travel and days off, too, and it marks them on the calendar.

It builds the week's sessions from your training history, recovery and schedule, then puts them on your calendar. You do not need to search a workout library. A different session starts with a message: even "endurance" is enough. You can ask for a particular interval structure or make any workout harder or easier in the chat.

A week in RestOrTrain

Monday: explaining pain on climbs

My knee hurts when I climb.

It remembers that detail and adjusts the week with your approval. It makes the change as soon as you tell it, explains its decision and keeps the pain in mind afterward.

It considers the message alongside your sleep, heart-rate variability (HRV) and resting heart rate. Your training load and power profile give it context for the work you have been doing. It also draws on earlier ride notes and conversations, including what it has learned about your strengths and weaknesses. It puts those details in the context of your current plan and calendar.

Tuesday: opening the planned workout

Open the workout on your bike computer. It sends planned sessions automatically to connected Garmin Edge devices and Garmin watches, Wahoo ELEMNT, Hammerhead Karoo and COROS. In Zwift, the session appears on the home screen on its scheduled day.

For an indoor session, you can use its workout player on iPhone or Android. Connect a Bluetooth smart trainer that supports FTMS. In ERG mode, the player controls the trainer to keep you at the prescribed power through each interval and recovery. You can turn ERG off or change the intensity with the minus and plus buttons. It continues with your phone locked. A heart-rate strap is optional.

Afterward, rides sync back from connections including Garmin, Wahoo, Zwift and Rouvy. Recovery data also arrives automatically: Garmin, COROS and WHOOP supply sleep, HRV and resting heart rate. Apple Health supplies those measures on iPhone too.

Wednesday: adding what the ride file missed

The last two intervals felt much harder than the first ones.

It uses that feedback when planning later sessions. It learns from every interaction, including activity notes and the details you share in conversation. It keeps that information organized for endurance training and uses it in subsequent decisions.

You can also inspect the ride's power curve, intervals and time in zones. If it matched a planned workout, it gets a Workout Score out of 100, with an explanation.

Facts remembered from your chats appear in the Memory list on the Profile tab. You can read them, add an entry or delete one. Ride notes remain separate, and still inform its answers.

Sunday: looking at progress through a training phase

Show how my power at threshold changed during this build phase.

It draws a bar or line chart in the reply from the training data it has already calculated. You can ask about the result in the same conversation.

The conversation also covers preparation beyond the sessions themselves. Ask about fueling your long ride or pacing your race, and it gives recommendations using the same history it uses to plan your training.

How it judges training load

Language models tend to prescribe too little endurance training: weeks with too little work and key sessions that are too short. RestOrTrain found that conservative bias in its own tests. Its AI is tuned to make the right training call, with decisions informed by current endurance science.

It also has guardrails specifically for assessing and recommending endurance load. General-purpose models do not have those guardrails.

How other apps use AI

Features checked October 9, 2026:

• TrainerRoad uses its own machine-learning models to select and tune workouts in structured cycling plans and plays them on a smart trainer in ERG mode. It has no chat.
• JOIN builds a plan around your fitness, goal and available days, reshuffles it around missed or extra rides, and adjusts it using Oura Readiness or WHOOP Recovery scores. It has no chat.
• Athletica has adaptive plans and an AI chat that explains and suggests but, in its own words, "does not automatically modify your training".
• General-purpose AI assistants such as ChatGPT, Claude and Gemini answer training questions, read files and screenshots, and remember details across chats when memory is on. None sends workouts to a bike computer, watch or trainer on its own.

Frequently asked questions

RestOrTrain plans your season and puts today's session on your trainer or Garmin.

A few questions a month are free.

Published June 13, 2026, updated October 10, 2026