Dieser Beitrag ist auch verfügbar auf: Deutsch
I trained with athletica.ai for six months and used it to prepare for important races. How does the AI training system work and what are the advantages compared to other apps?
Anzeige: I was able to use athletica free of charge during the test phase. This had no influence on the content of my review. The article is freely written and exclusively reflects my personal experiences.
Basics
athletica is a US-based company with a team of sports scientists and programmers. There is a close connection to HIITscience – another company that focuses on the science and application of High-Intensity Interval Training. Paul Laursen and Martin Buchheit’s expertise is evident in their reference book, which I consider to be one of the standard works in this field:
Those principles—and the plans derived from them—serve as Athletica.ai’s foundation. And whereas HIITscience centers on theory and guidance, Athletica.ai translates that knowledge into day-to-day training.
At athletica, there are no secrets – according to their own claims. And that is comprehensible:
- from day one, you’ll have full visibility into the training plan – which continuously adapts to your development.
- the core principles of the plans are presented in the book and grounded in scientific evidence.
- the forum welcomes all questions — and in my experience, responses are consistently quick and helpful.
- there is a blog with detailed explanations on individual topics or new features
- each week, the podcast dives into conversations with experts and showcases key concepts built into Athletica.ai.
Development status of athletica.ai
The platform is in constant development. The company was only founded in 2019 and I first looked at an early stage of development in 2022/2023. It was still miles away from the current status.
A lean, ten-person team is still refining core functions while steadily adding new features.
Many of the exciting new features are first available to beta users before they are finally rolled out. However, any (paying) user can join the beta circle. Of course, I have done the same and therefore some of the features described here may still be in the beta stage.
However, the release notes clearly show how regularly updates are published here and how the platform is being expanded. Especially in the past few weeks, development has accelerated, with seven March releases to show for it. This process is noticeably fueled by the many very active users in the forum who provide helpful tips and point out errors.
Principles
3-stage training plan model
Separate base plans exist for beginners, intermediate, and competitive runners (based on the research results from the book), on which the personalized training plans are then based. You can also specify the time available per week as a guideline for planning.
The weekly structure of the respective plans is always the same and varies only slightly depending on the training phase. This means that Thursdays, for example, are always VO2max intervals. However, the exact number of repetitions and the total length changes from week to week. In a recovery week, the intervals are sometimes omitted completely.
7 training zones
athletica’s training intensity model is based on 7 zones, with the top two zones only describing speeds for very short sprints. Essentially, the system is based on the familiar 5 zones. That lets you carry over familiar concepts like ‘Zone 2’ without adjustment.
The zones are calculated depending on the current threshold value and are based on heart rate or pace for runners. Zones update automatically when the system detects new threshold data—or after the optional test week.
Training load
Here too, athletica relies on scientifically proven concepts and uses its own modification of the Banister model to describe fitness, form and fatigue. However, the algorithm also includes personal factors such as RPE or feeling and also evaluates the comments in the notes field.
Here the system’s smarts show: if an individual session differs from plan, later workouts adjust accordingly.
And it’s fully transparent: you still see the original targets, plus an explanation of each adjustment.
Training targets are set either according to heart rate or pace. In intelligent mode, athletica automatically decides which metric is most appropriate for the workout in question. Basic runs are then controlled by heart rate, everything else by pace. The proper integration of power (e.g. via the Stryd) is still being worked on (as of 06/2025).
Starting with athletica.ai
Boundary conditions
- Athletica is English-only for now, and additional languages don’t seem imminent.
- The platform can be used via the website. However, there is now also an app with a limited range of functions.
- To be able to use all functions, athletica should be paired with a Garmin watch. However, other manufacturers can also be integrated using a detour via Intervals.icu.
Setting up the training plan
Athletica offers a two-week free trial, giving you time to put it through its paces.. After that, use costs 20$ per month.
When registering, you are guided through a dialog to enter the necessary basic information. This naturally includes some personal details (user name, password, e-mail, age), but the second step already includes the links to other platforms.
Best practice: hook up Garmin and/or Strava right away. This would give athletica a good basis for the last six weeks of training to build on with the training plan.
Then, of course, you are asked which sport you want to train for. athletica has a wide range of options here and can create training plans for triathlon, duathlon, running, cycling or rowing. The latest addition is HYROX. You can then set your experience level from beginner to professional in three stages and the training hours available per week.
If you know your threshold values, you can enter them on the next screen (pace and heart rate for running) or let the system determine them as part of a test week. If you know your values, you can skip the test week for the next query.
In principle, athletica can plan the entire season with several competitions, but you can also start training without a specific goal. However, you do need to specify what basic goal the plan should be based on (10k, half marathon, …).
You should then give the system enough time to import and evaluate the training history. As long as the arrows at the top right are still turning, the AI is working in the background. Nevertheless, you can of course already orient yourself on the platform.
Support
It makes sense to first take a look at the“Help & Education” section (drop-down menu top right). I can only recommend taking a look at the athletica Course to get an overview of the principles and functions. There is also a link to an e-book in the menu, which no longer quite corresponds to the program version, but can serve as a reference for important basics such as the zone model.
I can’t praise the forum enough; it’s become a regular part of my routine. If you have any questions when getting started with athletica, you can look them up here or ask your question. If it has not already been answered by other users, the athletica developers are guaranteed to get back to you quickly with an answer or further questions. This ranges from very basic questions about training to very specific problems or individual function requests. You’re never left hanging—the developers are engaged, open and transparent about technical issues.
Structure and functions
Overview
When navigating the website for the first time, you should start with the four tabs at the top center of the interface. At the top of the overview page you can see five pieces of status information for the current day. Below this is the graph of the “Performance Potential”, i.e. the load calculation according to Banister (with adjustments by athletica).
Below this you will find the current division of the zones for information purposes and, to the right, the long-term training plan with details of the weekly blocks as part of the periodization.
Training plan
The second tab shows the training plan as a weekly overview. An area can be displayed on the left that shows, for example, the total training kilometers for the week or warnings from the AI regarding the plan.
The current periodization step is always indicated above the weekly plan. I can immediately tell I’m in the build phase of a mid-volume 10 k schedule. So you can recognize built, taper or recovery phases just like the change between a 10k or half marathon plan.
Each training session contains an explanatory text on the purpose of the session and a brief description of the individual sections (if available) with details of the specified zones. Above the description, four icons lead to detailed information as a video or audio stream (Spotify, iTunes, YouTube and Captivate). This is particularly helpful for more complex workouts.
You can use the buttons to move through the weeks or jump back to the current week. However, only the next 1-2 weeks are based on the individual AI planning, which takes past loads into account. Further down the line, the plan initially remains generic (and highly repetitive). Yet it still gives an excellent long-term overview—something most AI platforms lack.
Charts
A whole series of graphical evaluations are summarized under the Charts tab. First and foremost, you can see the“Performance Potential“, i.e. the representation of the load control.
There are also pace and power profiles for cycling, running, swimming and rowing. The recovery profile deserves special attention, as the daily HRV measurement is logged here and conclusions are drawn from it.
Athletica.ai can work with different sources. Night-time measurement with Garmin watches is directly supported. However, values can also be transferred from intervals.icu, which opens up further possibilities. I send my morning measurements with HRV4Training (manually) there and thus also have the values in athletica.ai.
Season planning with athletica.ai
If you want to prepare for a specific competition or an entire season with Athletica, you can enter the planned events in the settings under “Goals”. In addition to the title, at least the date and the type of competition (10k, half marathon, marathon, …) must be entered. The priority is also mandatory and influences the length of the taper before the race day.
Manual interventions
The advantage of an intelligent training plan like athletica’s is, of course, that it can respond to manual interventions at any time. Training sessions can be moved in the calendar using drag & drop to adapt them to personal time management. If necessary, other workouts then react to this adjustment.
The workout wizard is also a particularly powerful feature. Here, the AI trainer provides alternatives to the planned workout. If you are short on time, you can choose interval training instead of an endurance run. Or swap interval training for an endurance run if you are still too exhausted from the last session.
You can also break out of the very repetitive structure of the training plan and choose a slightly different version of the session with the same training effect. If you are injured, you can choose from a range of alternatives, from swimming or cycling to stretching or yoga. And finally, you can simply switch sports and get the training stimulus on the bike, in the swimming pool or on the rowing machine.
Advanced features
Athletica.ai also offers other features that you might not notice at first glance. For example, you can enter your availability on the individual days of the week as the basis for the training plan. For example, I always do Taekwondo training on Tuesdays and I don’t want Athletica to schedule this day. So I have specified a rest day here.
I’m housebound on Fridays, but I could of course do strength training or indoor rowing. Since I rarely have more than one hour, I can specify that in the settings. The system also warns you if your constraints are too restrictive and suggests what to revisit.
The library of your own or preconfigured training sessions is also not immediately obvious. It can be accessed via the left sidebar in the training plan view. Here you can store your own workouts that you want to add manually to the weekly plan if required. Otherwise, you can access all standard Athletica sessions via the Global Library.
Connection of further platforms
Strava
Connect Strava once, and your full activity history populates Athletica automatically. This data forms an important basis when you start using the tool. But I also find the sync very helpful down the line, because it means that the AI also receives workouts that I don’t record on a watch, for example (taekwondo, yoga, etc.). Athletica then also takes care of naming the session if it was in the training plan.
Garmin
When I started with athletica.ai six months ago, only Garmin watches could be integrated. You benefit from the sync with the platform by transferring planned training sessions, which you can then carry out with guidance – including pace or heart rate specifications for the individual sections. Conversely, the recorded data is then sent back to Athletica.
Another advantage is the transmission of night-time HRV measurements if you wear your watch 24/7. Athletica is one of the few platforms that can derive decisions for training management from this. However, I’m not entirely sure whether the results really interfere with the AI. However, their assessment is always communicated in the weekly report or the recovery profile.
To be honest, this is still the ideal solution, even if other options are now available.
Intervals.icu
I was particularly pleased to hear that Intervals.icu has been fully integrated. This has been my training diary of choice for some time now and has completely replaced TrainingPeaks. In combination with athletica.ai, it opens up completely new possibilities. Because Athletica exports its entire training plan there, the units can also be transferred to all watches that Intervals.icu supports – including Suunto, Coros or Polar. Their recordings would of course also end up back at Athletica via Strava, but there is now another option via Intervals.icu.
The connection is also a big advantage for me because I no longer have to rely on the HRV measurement from Garmin. I’ve been measuring every morning with HRV4Training for years and naturally wanted to keep it that way. As this data ends up on Intervals.icu anyway, it is now also available to Athletica. Perfect!
Wahoo and Concept2
Athletica.ai is not only aimed at runners, however, so there are other connections, for example to Wahoo. Their bike computers and roller trainers are also widely used and can be connected directly via Athletica.
Athletica.ai is also probably one of the few platforms that can create AI training plans for rowing. Here, too, there is a sync with the logbook from Concept2, the leading provider of indoor rowing machines.
My experience with Athletica.ai
Everyday training
I’ve been using Athletica.ai for six months now and have really worked with it. And I went all-in – no backup plan on any other platform.
Transparency
I really liked the fact that Athletica plans the entire season in advance and that I can always see which training phases the system has planned via the weekly blocks. There is a specific plan with all the workouts for each week and I was able to easily check, for example, whether the parameters selected for the training plan led to a sensible result. You only have to look at the weekly kilometers or the number and content of the sessions. This has helped me a lot to optimize my targets.

I ended up at the “Recreational” experience level and 4-6 hours per week, which leads to a mid-volume plan. This in turn means 35-50 kilometers per week for me, whereby the standard week is around 45 kilometers (in preparation for a half marathon). The 10k plan weeks are slightly less.
Weekly structure
The weekly structure is always the same and that confused me at first. Do I really want to do 30/30 HIIT intervals every Thursday? My experience after the first few weeks was: yes, I do! The consistency makes it far easier to weave training into daily life. It is also good for the mind to know on which days you have to prepare for a higher workload.
It also makes it possible to compare the sessions and thus gain a better insight into your own training progress. Just last week I noticed that I am now running the 30/30 HIIT intervals 15 s/km faster than four weeks ago.
In the beginning, the explanations on Spotify or YouTube for the individual sessions also helped me a lot to better understand the content and the ideas behind them. Especially with the quality sessions, of course. But that was only a very short phase, because the structure soon becomes second nature thanks to the repetition.
Load management
The load management also works very well for me. At first I wasn’t so sure whether I could do the weekly volumes regularly. And to be honest, I sometimes shortened or skipped a session – especially in the first few weeks. But the adaptation took effect very quickly and the plan never overwhelmed or underwhelmed me. After all, recovery weeks are also regularly scheduled.
In the past six months, I have trained more than I have for a very long time (volume, intensity and frequency). Despite this, I wasn’t injured or ill. My HRV was stable. It’s fair to say Athletica nailed the load management.
Supplementary training sessions
In addition to pure running units, my training plans always included supplementary workouts. Initially, these were cycling sessions that replaced basic runs. In the more specific preparation for competitions, this then became strength and stability training, which also included plyometric exercises. These in particular have really pushed me forward!
But to be honest, sometimes these sessions also bothered me and I would have liked to get rid of them. Unfortunately, this is not possible by default. But you can of course intervene manually at any time, delete the unit and replace it with another one from the library. Or simply take a break for a day.
AI adjustments to the training plan
It only took a few weeks before I noticed for the first time that the AI was adapting to me. Because even before you could set the availability on weekdays, Athletica quickly noticed that I always postponed the Friday sessions manually and then planned a rest day instead.
I was also able to quickly understand the adaptation of the AI, especially in the HIIT units. Initially, the training plan always included the same unit, but this was soon varied. 1x5x30s then became 1x8x30s or 2x6x30s and so on. Of course, this also applies to the other training content, but it is more obvious with the intervals.
After each completed session, there is also feedback from the AI coach, who praised me for sticking to the zones well, for example, or sometimes reminded me to stick to them better. Basically, however, the developers of Athletica want a “responsible athlete” who uses the system as a tool but makes their own decisions. The AI is designed to adapt to you – not the other way around.
Dealing with seasonal planning
As mentioned above, Athletica can not only deal with a specific race goal, but also plan the entire season. In my case, this meant planning the entire Westmünsterland running series in addition to the Enschede half marathon in the spring.
The running series consists of 10 races, some of which are very close (one week apart). On the one hand, I had to set clear priorities, as of course not every run can be an A race. But even if you work with B and C priorities, there were too many taper and recovery weeks in my training plan.
That’s why I only gave the AI coach the A competitions and did the other runs out of training. In each week, I put the quality session with the highest load on the day of the competition. You shouldn’t underestimate the AI here, because of course it wants to reschedule immediately and restore the original plan.
What I noticed in the actual race preparation for the half marathon: the plan doesn’t include the usual long runs at the weekend. According to Athletica, the longest distance in training would have been around 16 km. This was done deliberately, because with 4 training sessions a week and around 50 kilometers per week, you assume that you don’t need the really long run.
Competition results
Whether all of this leads to the desired results is something you notice on race day. For me, the first “day of truth” was the half marathon in Enschede, for which I spent four months preparing with Athletica. It was only last year at the Cologne Half Marathon that I managed to stay under 1:45h again after years. I even stayed just under 1:44h and was therefore hoping to run 1:43h in Enschede – if things went really well. With 1:41:50, I more than beat my desired goal and ran my best half marathon so far (corrected for age)!
I was also able to improve or at least match my personal best times in the subsequent 10 km runs. These are times that date back years in some cases, and I didn’t think I would ever be able to achieve them again. With the help of the Athletica training plan, I have closed this gap with a huge leap forward.
Stumbling blocks
No system is without its flaws. But in the case of Atheltica.ai, this seems to be mainly due to the fact that the platform is relatively young and there simply hasn’t been time to tackle these issues yet.
Manual interventions in the training plan (workout deleted or moved, event added, threshold values changed, …) trigger a recalculation, which is usually completed after a few moments, at the latest after a few minutes. Sometimes I notice that although the next sessions have been adjusted appropriately, the following week has some oddities. These are usually corrected after about 12 hours of waiting (overnight), but until then this naturally leads to irritation.
The handling of competition days could also do with a few more programming lessons: Athletica only allows competitions to take place on a weekend and they cannot be moved to the week – for example, if an event takes place on a public holiday. In this case, the “responsible athlete” is required… The AI doesn’t seem to recognise the competition day itself as a planned unit. Because on the following day you’ll get flagged for an unplanned spike in workload…
What should actually be possible as part of the beta functions is training by power. Even if the platform displays meaningful power values for the sessions, the structured training workouts at Garmin are still based on pace or heart rate. However, implementation is surely only a matter of time.
Support and development speed
In general, the speed of development has been very high, especially in the last few months. So fast, in fact, that I had to adapt my article several times during its creation in order to keep it up to date. New functions are first made available to beta users, but then very quickly made available to all users.
You can become a beta user by making a simple enquiry in the forum, which is my almost daily point of contact. It seems that the entire Athletica team is represented here, from co-founder Paul Laursen (@prof) to Marjaana and the developers. Many dedicated and experienced users also ensure that all questions can be answered within a very short time. And if this is not possible, for example if you ask about a function that is still being developed, you will always get a very transparent answer. The forum support is genuinely outstanding.
Comparison with other platforms
I have been training exclusively according to AI training plans from various platforms for three years now and have generally had good experiences with them. However, there are clear differences in direct comparison.
athletica.ai vs Enduco
I’ve known the German start-up Enduco and its training plans since 2022, and they have a lot in common with athletica.ai:
- Training plans for runners and cyclists
- Complete season planning possible
- Good load control
- Consideration of the time frame per weekday
- Consideration of subjective stress (RPE)
Transparency is Athletica’s standout advantage for me. Unlike Enduco, I can see the entire training plan for the whole season, not just the next week. The basics of the training plan are also much easier to understand in Athletica thanks to the book.
And even though I really like the Enduco app, I find a website like Athletica’s more convenient to use.
But the strongest argument is definitely that I had much better success in competitions by training with atheltica.ai. Even though the results with Enduco were always solid, I never made as much progress as with Athletica.
However, at €10 per month (for a long-term subscription), Enduco costs only half as much.
athletica.ai vs TRAIT (former TWAIV)
I have also had good training experiences with TWAIV. It is based on well-founded training plans, but again these are not made transparent. However, you can view the entire plan – even if at some point only the key workouts are displayed.
The biggest difference for me is the focus of the two programmes: In my opinion, TRAIT is primarily aimed at beginners and has a lot of good ideas to keep you going. The information on offer, the price structure and the additional exercises are also clearly aimed at this target group.
Athletica’s training plans are just as useful for beginners, but they do require a little more background knowledge. Although this knowledge is also available here, you are not guided through it as closely. Athletica.ai is definitely at its best when used by ambitious amateur runners or professionals.
Conclusion
As you can see, I have worked extensively with athletica.ai and gained experience with it for half a year. This includes, above all, competitions with new best times, which is certainly the best basis for assessing training plans. 😉 In this respect, Athletica has fully convinced me and will continue to plan my training in the future.
At the same time, athletica.ai is not perfect. Minor technical problems crop up from time to time and larger components such as the free choice of competition days or training by power have not yet been implemented. So of course you have to weigh up whether you can accept this with a paid service, which at $20 per month is not exactly cheap.
For me, however, the equation works above all because there is a very committed team behind it and there is very regular progress in the development of athletica.ai. The support via the forum is also exemplary, so that small stumbling blocks were always easily overcome.
Most importantly, I’ve never trained as effectively as I have with Athletica.ai! The training can be ideally adapted to my goals, my performance level and the time available – and responds to deviations and manual adjustments at any time. I was always well challenged, but never overwhelmed, and was optimally prepared for all competitions. Last but not least, I really enjoy training with Athletica. 😀