A useful definition
The coach gives the guidance. AI helps keep it current.
RunPT is a personal running coach that builds a phased plan, tells you what to run, and explains why each workout matters. Its AI coaching system uses your goals, constraints, and training data to review what comes next and answer questions about the plan.
The useful part is not the word AI. It is the ability to keep multiple pieces of evidence in the same planning decision: what you are training for, what you intended to do, what you completed, what recovery context is available, and what the current phase is meant to accomplish.
The inputs
What RunPT can consider.
Goal and timing
Target distance, race date, experience, current volume, and recent reference performances establish direction and a realistic runway.
Weekly constraints
Available run days, rest days, long-run preference, strength work, and other commitments define what can fit safely into a week.
Completed workouts
Synced runs provide evidence about duration, distance, intensity, execution, recent consistency, and the load already absorbed.
Recovery context
Available sleep, HRV, resting heart rate, readiness, and stress can add context when compared with a runner's personal baseline.
Training phase
Maintain, Base, Build, Peak, Taper, and Recovery give each session a purpose inside the wider block.
The conversation
Questions, scheduling constraints, and feedback can help the AI coach explain or contextualize the plan without turning chat into medical advice.
Possible plan changes
RunPT can adjust upcoming work without treating change as the goal.
When the evidence materially changes what should come next, RunPT can alter the placement, intensity, duration, or type of an upcoming session, rebalance the week, or reconsider progression. If the plan remains valid, it can stay in place.
RunPT does not claim that the whole calendar rebuilds after every run. Newly synced workouts are evaluated, and meaningful updates are made when the planning case supports them.
- Preserve valid sessions when no useful change is needed
- Respect declared rest days and the chosen long-run day
- Avoid automatically moving missed quality work into the next day
- Keep the current phase and race timeline in view
Conversational coaching
Ask about the reason, not just the instruction.
A training plan is easier to follow when the runner can understand why a session is there. You can use the RunPT coach to ask what today's workout develops, how a busy week affects the block, or what recent training suggests.
Responses can use relevant profile, plan, workout, and available recovery context. They are generated by AI and can be wrong. Important health, injury, and high-stakes training decisions still belong with appropriately qualified professionals.
Illustrative scenarios
What adaptation may look like.
An unexpectedly hard long run
If duration and intensity create more load than expected, RunPT may keep the next day easy and reconsider nearby quality work. It may also leave the week unchanged when the broader evidence does not support an update.
A missed interval session
The next planning pass can consider whether the session still matters, whether it fits elsewhere, or whether the week is better without it. The default is not to stack work simply to restore a perfect calendar.
Several signs of reduced recovery
A combination of changed sleep, HRV, resting heart rate, and recent load may support easing or moving upcoming work. Those signals are still considered against the runner's baseline rather than treated as a diagnosis.
Faster-than-expected progress
Stronger completed sessions and improving recent evidence may support a progression when it fits the phase. RunPT does not use one quick workout as proof that the entire plan should accelerate.
Race day is approaching
As the race date nears, the current phase matters more than chasing new fitness. RunPT can emphasize race-specific sharpening in Peak and protect readiness through Taper rather than adding late volume.
These examples illustrate the planning logic; they are not promises that the same input will always produce the same update.