.png)
Can ChatGPT Create A Marathon Training Plan? What Runners Need To Know
Artificial intelligence is changing how runners approach marathon training. Instead of downloading a fixed training schedule or paying for a personalised programme, runners can now ask AI tools to create a plan based on their experience, goals and available time.
ChatGPT is one of several AI assistants capable of generating training schedules, alongside alternatives such as Google's Gemini and Anthropic's Claude. Although these systems work differently and may produce different recommendations, they offer similar opportunities for runners to request personalised training advice. For this article, we're using ChatGPT as our example, but many of the questions we explore are equally relevant to other AI tools.
But how good are the plans these systems produce? Can they account for the differences between a first time marathon runner and someone chasing a personal best? And can they recognise when a training schedule might be unrealistic or place too much demand on the runner?
How We Assessed ChatGPT's Marathon Training Plans
We created three fictional runners representing common challenges during marathon preparation. The first was preparing for their debut marathon, the second wanted to improve an established personal best, and the third needed to fit training around a demanding schedule.
For each profile, we examined whether the suggested programme reflected the runner's circumstances, included appropriate recovery and developed training gradually. We also explored whether providing more information could improve the relevance of the recommendations.
The schedules were illustrative examples generated within ChatGPT for this article, rather than independently tested programmes followed by real runners. Our observations therefore highlight features of these particular examples, not how reliably AI systems perform across different attempts.
Test One: The First Time Marathon Runner
Our first runner was 35 years old and preparing for their first marathon, in 20 weeks. They currently ran three times a week, covering approximately 20 kilometres in total, with a longest recent run of 10 kilometres. Their goal was simply to finish, and they could train four days a week.
ChatGPT produced a programme that started at 20 kilometres per week and gradually introduced a fourth running day. Weekly mileage increased to a peak of 48 kilometres, with recovery weeks and a reduction in training before race day.
The schedule contained many familiar elements of beginner marathon preparation, including easy running and progressively longer weekend sessions. However, one recommendation stood out: the longest run increased from 20 kilometres in week 15 to 30 kilometres in week 16.
That represented a 50% increase in the distance of the session, while the 30 kilometre run accounted for more than 60% of the week's total mileage. Although the preceding week involved reduced training, the size of the increase deserved closer examination.
The example exposed a serious weakness in ChatGPT's training recommendations. Increasing a beginner's long run from 20 to 30 kilometres in a single week is an excessive and poorly judged progression that could significantly increase fatigue and injury risk. This is not a minor detail in an otherwise sensible programme, but a fundamental flaw in how the training has been structured. It demonstrates why runners should not assume that an AI generated marathon plan is safe or appropriate simply because it looks convincing.
Test Two: The Experienced Runner Targeting A Personal Best
Our second runner was 42 years old and had completed five marathons. Their personal best was 3 hours 35 minutes, and they wanted to improve that to under 3 hours 20 minutes.
They were already running approximately 55 kilometres a week and had recently completed a half marathon in 1 hour 36 minutes. With 16 weeks until their next marathon, they could train five or six days a week.
ChatGPT generated a more demanding programme, increasing weekly mileage to a peak of 78 kilometres. It included threshold sessions, marathon pace running and long runs of up to 32 kilometres, alongside easier running and recovery periods.
This showed that the system could distinguish between preparing a beginner to finish and helping an experienced runner target a faster time. However, the programme regularly included threshold running on Tuesday, marathon pace work on Thursday and a demanding long run on Sunday.
That combination could be appropriate for some runners, but the schedule did not provide enough detail about the faster sessions to judge the overall workload. It also accepted the proposed finishing time without adequately examining whether the runner's recent performances supported it.
The main concern was not the presence of demanding workouts, but whether ChatGPT had provided sufficient information to establish that they were suitable for this particular runner.
Test Three: The Runner With Limited Training Time
Our third runner was 39 years old, worked full time and had family commitments. They had completed one marathon previously and currently ran around 25 kilometres a week.
They had 16 weeks until their next marathon but could only run three days a week. Their priority was to finish comfortably without allowing training to dominate their schedule.
ChatGPT respected the three day restriction, creating a programme built around two shorter weekday runs and a longer weekend session. Weekly mileage increased to a peak of 43 kilometres, with a longest run of 28 kilometres.
The schedule appeared practical because it fitted around the runner's availability and included recovery days. However, the 28 kilometre long run accounted for approximately 65% of the week's total distance and followed an 18 kilometre long run the previous week.
This raised questions about how the programme balanced marathon preparation with limited training time. Rather than reconsidering the progression, the schedule appeared to prioritise reaching a conventional long run distance within the available weeks.
The example highlighted an important limitation: adapting a programme to fewer training days does not automatically mean the resulting workload is appropriate.
Does Giving ChatGPT More Information Improve The Plan?
One of ChatGPT's potential advantages is that runners can describe their experience, current fitness and availability rather than choosing from a fixed collection of training schedules.
To explore this, we compared two illustrative requests for a beginner marathon programme. The first simply asked ChatGPT to create a 20 week plan, while the second explained the runner's current mileage, longest recent run, available training days and objective.
The basic request produced a generic schedule starting at 24 kilometres per week, with a longest run of 32 kilometres. The detailed request began at the runner's established 20 kilometres per week and introduced an additional training day gradually.
Providing more information helped the recommendations reflect the runner's circumstances. However, the detailed schedule still contained the questionable long run progression identified in our first example.
This illustrates why runners should provide accurate information when requesting a plan, while recognising that greater personalisation does not guarantee better training decisions.
What Should Runners Check Before Following An AI Training Plan?
Our examples highlighted the importance of looking beyond the overall appearance of a training schedule. A programme may contain familiar workouts, recovery weeks and a taper while still including individual sessions that deserve questioning.
The starting workload should reflect what the runner is already doing, rather than assuming they can immediately manage a substantial increase in training. Weekly mileage and long run distances should develop in a way that accounts for experience, consistency and recovery.
Runners should also consider how demanding sessions are distributed throughout the week. Several harder workouts may be manageable for an experienced athlete, but the overall workload needs to reflect their fitness and previous training.
Practical circumstances matter too. A schedule that technically fits around work and family commitments may still create considerable fatigue if too much running is concentrated into a small number of sessions.
If a recommendation appears unrealistic, runners can ask ChatGPT to explain its reasoning or suggest alternatives. However, a convincing explanation should not be mistaken for independent confirmation that the programme is suitable.
Can ChatGPT Replace A Marathon Running Coach?
ChatGPT can provide useful support beyond generating a training schedule. Runners can ask it to explain different types of workouts, explore training principles and help organise practical aspects of marathon preparation.
It can also respond when circumstances change. Someone who misses training because of work commitments could request suggestions for adjusting their programme rather than attempting to complete every missed session.
However, ChatGPT relies on the information supplied by the runner. It cannot directly observe fatigue, assess physical condition or independently establish whether someone is recovering adequately.
An experienced running coach can discuss how an athlete responds to training, identify developing concerns and adjust the programme based on observations over time. AI can respond to new information, but that is not equivalent to ongoing individual coaching.
For runners without access to a coach, ChatGPT may still be a useful source of training information. Its greatest value may be helping people understand their preparation and explore options, rather than replacing professional judgement.
Final Thoughts
Our illustrative exercise showed that ChatGPT could generate recognisable marathon training plans for runners with different experience levels, goals and time constraints.
The schedules included long runs, recovery periods and adjustments to training frequency.
However, closer examination revealed questionable increases in long run distance and insufficient detail about some demanding workouts. These weaknesses were not immediately apparent from the overall structure of the plans.
Providing more information helped ChatGPT produce recommendations that better reflected the beginner's circumstances, but it did not eliminate every concern. A personalised looking programme is not necessarily an appropriate one.
AI tools such as ChatGPT, Gemini and Claude can make marathon training information more accessible and help runners explore their options. The important consideration is how that information is used, with realistic expectations, critical judgement and appropriate professional advice when needed.
Comments