AI can create your training plan. But can it coach you? A practical guide to creating effective training plans for self-coached runners.
Until a few months ago, Vinny Mauri was largely unknown even within running circles. Then, in April 2026, he ran 2:05:54 at the Glass City Marathon in Toledo, becoming the fastest American debut marathoner ever. It was the same day Sebastian Sawe broke the two-hour barrier in London. Mauri’s run was overshadowed by that historic feat, though he featured in GeeksOnFeet’s New Rules of Marathon Training podcast, recorded just days later.
The unusual part of Mauri’s story was his training. Mauri is self-coached, and ran most of his mileage on ordinary Planet Fitness (gym chain in US) treadmills, logged a single-day 45-mile run and shared his training on Strava. His non-traditional approach to training made the 25-year-old an unlikely media sensation. Wall Street Journal profiled the “breakout marathon star” and his unconventional methods.
During the Chicago Marathon, Mauri surged at 15K to open a gap on his fancied American teammates and stayed on American-record pace until halfway. Hamstring problems forced him to stop several times, but he still finished in 2:06:45, tenth overall and third American: a noteworthy performance on a world-class stage.
Mauri’s story raises a question for the rest of us: how should a self-coached runner decide what to do, how much to do, and when to push or recover?
While Mauri did not use an AI coach, current AI assistants are very capable, and are easily accessible to self-coached runners. AI tools offer a compelling proposition: enter running history, target time and availability, and get a detailed schedule of workouts, paces, long runs and taper weeks. The result can be very convincing. But does it identify the right priorities?
Suneel Kumar Koyi, a seasoned Bengaluru runner with seven sub-three-hour marathons, used Claude to build his Bengaluru Marathon cycle. He supplied his training history from Garmin and Runalyze, key Strava workouts and a sub-three-hour goal, then followed the plan closely: 1,470km across 103 runs, including 100 running days, a 38km long run and 127km at marathon pace. He finished the marathon in 2:59:34, his slowest sub-three-hour marathon in seven attempts.
In a thoughtful post-mortem on X, Suneel identified what he saw as a key gap: the AI plan averaged just 4.6km of threshold (tempo) running per week, compared with 12.4km in his best previous build.
A fun fact about my Wipro Bengaluru Marathon 2026 training:
— Suneel Kumar Koyi (@suneelkoyi) October 6, 2026
The entire training plan was built by Claude.
Goal was around 2:55 to 2:57
No human coach reviewed it.
I followed it religiously. pic.twitter.com/MJ0mnO78Sd
His conclusion was telling: AI can produce a convincing plan, but the critical eye of a coach can question whether it suits the runner’s abilities, history and goals. For his next training cycle, Suneel is using RunStrong with a carefully crafted custom plan, while also seeking input from experienced runners and coaches. He believes AI plans may be sufficient for beginners and intermediate runners, but more competitive athletes benefit from expert feedback.
The lesson from Suneel’s experience is broader than whether AI training works or not: any plan deserves scrutiny, especially when it departs from a runner’s successful training history.
A good training plan is a sequence of decisions designed to produce a specific outcome. Its quality rests on three principles.
Specificity. Training must reflect the event. A 10K runner needs to sustain relatively high intensity; a marathoner must combine aerobic capacity with endurance, durability and fuelling over several hours. The qualities overlap, but a plan suited to one distance may leave gaps for another.
Individual variance. Two runners targeting a sub-three-hour marathon may need different plans. One may have speed but lack endurance; another may have a strong aerobic base but struggle at threshold. Training history, recent performances, injury record, mileage, available days and recovery capacity all matter. Chronological age is only one factor; running age and tolerance for training load are equally relevant.
Stress + recovery = growth. Training supplies the stimulus; recovery allows adaptation. More work helps only up to a point. A good plan balances hard sessions with easy running, rest and lighter weeks so fitness can build without unsustainable fatigue.
Legendary coach Jack Daniels captures the principle in Daniels' Running Formula: “You should always be able to answer this very simple question: ‘What is the purpose of this workout?'” Runners may do intervals, threshold sessions and long runs without knowing what each is meant to develop. Understanding the purpose of each run is the first step towards a sound training structure.
Start with the goal race, current fitness and time available. Work backwards from race day, but build forwards from what the runner can handle today. Most plans move through three overlapping phases:
| Phase | Primary objective | Typical training |
|---|---|---|
| Base | Build aerobic capacity and prepare for higher loads | Easy mileage, long runs, strides, hill sprints and short 200–400m repetitions |
| Development | Improve qualities relevant to the event | Threshold sessions, 3–5-minute aerobic intervals, longer runs and progressive mileage |
| Race-specific | Convert fitness into race readiness, then shed fatigue | Goal-pace blocks, specific long runs, fuelling practice, sharpening and taper |
The phases are not sealed compartments. A marathoner might build towards 2K–4K threshold repetitions, then add marathon-pace segments to long runs; a 10K runner may focus more on race-pace intervals. Easy running remains important throughout.
Periodisation gives the cycle structure: macro-periodisation organises the full build, while micro-periodisation balances hard sessions and recovery within each week. The familiar 80:20 easy-to-hard split is a reference, not a rule for everyone. Three building weeks followed by a lighter fourth can help manage fatigue, but the right rhythm depends on the runner and the load. Aim for consistent progress, not an awe-inspiring week that compromises what follows.
Runners have three broad routes to a plan: hire a coach, learn to coach themselves, or use an online or AI-assisted tool. These can complement one another.
A good coach brings experience, pattern recognition, accountability and the ability to interpret training responses. They can spot accumulating fatigue, adjust a workout and help reconsider a goal. Self-coaching can work well for runners willing to study the principles and assess progress honestly, using established books or AI tools as foundations. Whatever the tool, personalisation and feedback matter.
AI can explain workouts, compare approaches and organise training around work and family. Treat an AI-generated schedule as a draft: check progression, race specificity, recovery and intensity distribution. Compare it with established principles and your own history, and seek a review from an experienced runner or coach where possible.
How to use AI effectively
Connect your data first. AI assistants like Claude and ChatGPT are most effective when they have full context about your training and goals. Instead of cutting and pasting, connectors built on MCP (the Model Context Protocol) are the best way to give them that:
Validate the plan. With RunStrong connected, ask the assistant to read RunStrong’s plan structure guide and review the draft against it: the purpose of each phase, long-run progression, spacing between hard days, down weeks and the taper. Then compare the key quantities with your best previous cycle. For example, threshold kilometres per week across the build, which is where Suneel found his gap.
Validate the load. With Strava or COROS connected, ask for weekly volume, the easy-to-hard split and the week-on-week change, and have it flag any week more than roughly 10% above the previous one or any run of three hard days without an easy day between. In RunStrong, ask it to compare planned with completed for each week and list what was missed. Community skills exist for session-RPE and acute-to-chronic load ratios; treat their thresholds as a starting point, not a verdict.
And always ask the Daniels question. For any session you cannot explain, ask what its purpose is. If the assistant cannot answer from your own data and history, that is the gap.
A skill that runs the plan checks for you
RunStrong has packaged the plan checks above as a skill for Claude and ChatGPT, the Training Plan Validator. Give it your plan, a recent race time and your last four weeks of training. It audits the starting volume, progression and lighter weeks, the easy-to-hard balance, hard-day spacing, how much threshold and interval work sits in one session, the long run and taper for your distance, and whether the paces come from your fitness rather than a goal time, then reports what passes, what fails and how to fix it.
Get the Training Plan Validator from RunStrong, with setup notes for Claude, Claude Code and ChatGPT.
| Approach | What it offers | What to watch for |
|---|---|---|
| Human coach | Experience, feedback and accountability | Cost, quality and fit |
| Self-coaching | Autonomy and deeper understanding | Knowledge gaps and personal bias |
| AI / online tools | Accessible structure and rapid planning | Convincing plans with important gaps |
| Hybrid approach | Technology plus human judgement | The runner must still evaluate and adapt |
Mauri’s rise shows that unconventional training can produce extraordinary performances. Suneel’s experience shows why even a plan followed faithfully deserves scrutiny. So the takeaway is there is no universal formula, but stick to essentials.
For self-coached runners, the essentials are straightforward:
The essentials
We are all experiments of one. Understanding training principles helps us interpret those experiments, whether the plan comes from a coach, a book or AI. Technology can make planning easier, but the ability to recognise a gap, question a recommendation and change course remains a source of sustainable competitive advantage.
RunStrong A plan you can question, built on your own history See the training plans →Related in this column: Pace the Race — pacing is budgeting, The Fuel Switch — metabolic flexibility for endurance runners and Into Thin Air — what altitude teaches every runner.
