This weekend I saw Chris from outdoorbros mentioned he followed a chatgpt plan for this years Latoja 200mi race.
I never thought about that and decided to ask for some different plans and I was actually pretty impressed with what it came up with! It seems to manage and build decent stuff for in season, off season, race prep goals, etc. both cycling and strength!
this seems to put an end on “buying” plans if you won’t have constant and direct access to a coach aka pre-built plans
Here’s 2 examples, this is just the summary there is a lot more detail it provided
Ride 2 (Thu/Fri) – Secondary interval day (different focus than Ride 1)
Ride 3 (Sat) – Long endurance ride (3–4h Z2, add tempo blocks later)
Ride 4 (Sun) – Steady tempo/Z2 endurance (2–2.5h)
Mon & Fri (if Thu is riding day) – Rest or optional 45–60min recovery spin.
You want an in-season strength training program for cyclists, with 2 sessions per week over 16 weeks. Since this is in-season, the goal is maintenance and injury prevention, not heavy strength gains. The key is low volume, moderate load, and focus on mobility and stability so it complements your cycling instead of interfering with it.
Here’s a 16-week in-season strength plan (2x/week):
Program Overview
Frequency: 2x per week (e.g., Tue & Fri, away from hardest bike sessions)
I did (for cycling). And I’ve learned quite a few things. Or perhaps was reminded of them again as I knew them before. In the end I am back to programming myself. But was using Gemini’s plans for 3 months. The most important improvement is that I started using HRV and that led to a better feedback after training.
tl;dr Personally, I would use an LLM in two scenarios: either for a 10,000 m overview or to fine tune the details, e. g. a concrete workout or training week (microcycle).
I would only recommend you trying if you have a solid background in structured training. You need to be able to separate the wheat from the chaff. It is so easy to get answers, which sound plausible at first glance, but are complete non-sense.
The devil is in the details, and some of the training plans are ok from a 10,000 m view, when you zoom in, they don’t make sense. E. g. I asked Qwen (a local LLM) to generate a plan for a critical racer with experience in structured training. The answer sounds quite reasonable, but it completely misses any notion of progression, and typically contains one all-out workout per week (1-minute and 30-second sprints). Importantly, there were no regular rest weeks. That’s a recipe for burn out. Now you would probably get rest weeks after you add that to your prompt, but again, you have to know that in advance.
What I liked about LLM-generated suggestion is that it explained its reasoning and highlighted the importance of e. g. active recovery (with yoga) and strength training.
You may also use it to fine tune some aspects of your training. But I don’t think it’ll be useful to give you concrete workouts that are tailored to you and keep you in the right zone to achieve progressive overload.
I’ve played with it and a coach and sport scientist it generally has reasonable logic and approach. The issue I have is it can change its mind almost instantly with very small prompts. I’ve put in session data which was followed to a tee as it recommended and its completely shredded the session as inappropriate or whatever. So I think the consistency and continuity is a major issue. I think its good for initially planning things out and organising but expecting it to track along a programme and adjust is problematic.
I’ve used it myself, and as the others before me has stated, I would not trust it to give me a good plan, it can deliver a plan but there is almost always something lacking, and you have to know what you are looking for.
I made myself a 12 week strength training plan, made it use a specific strength training protocol (with link to it) as a template and asked for 12 weeks with progressive overload, and to make it into a excel sheet for printing. In that case it just does all the groundwork for me, takes the exercises i want to progress in and gives me a 12 week progressive overloading plan in a excel sheet, then i can fine tune it and leave some extra columns blank for notes etc.
The first revisions i tried to use chat gpt to do all the fine tuning, but even when given a small precise input it alters the whole plan.
Also, for me having a physical sheet to fill out takes away a lot of mental barriers and lets me focus on just doing what’s on the papersheet i have.
Not exactly a Training Plan, but I am experimenting with ChatGPT as a coach. I tell it how I feel, injuries, sleep, etc. and it tells me which Sport I should do (swim/bike/run) and if Z2 or intervals would be better. I give it feedback every morning and after each Session and it ajdusts.
The onboarding process was quite extensive, after I told ChatGPT, what I want and asked what it needs to do that for me.
Till now I am shocked how well it works. It feels like a superhuman coach that is always accessible.
I tried with ChatGPT several months back and it was an abomination. I had to tell it the same thing over and over again and it still never got it right. For example, I put a cap on hours and it returned a plan that exceeded the cap. I told it, it apologized and gave me another plan that exceeded it. I told it again, it told me I was right, apologized, and did it a third time! There were other major errors involved and I gave up. Sounds like maybe it’s improved since then based on the comments above.
Train, observe, adjust, see No, sorry, I don’t have more to add. Just repeatedly doing high intensity stuff I couldn’t sleep well, couldn’t recover well, was stressed before training and saw it in HRV for a couple of days or more. Changing to threshold training, felt better, recovered better and HRV reflected that as well. It might be superfluous as I had enough indicators already. But it is nice to have some objective data as well to see that you are not making thins up.
I’m currently using ChatGPT as a coach. It helps me plan my training, nutrition, recovery, and core work. The main goal is to build back up sustainably and get above 5 W/kg again, without overdoing it or burning out. I still listen to my body, but having a structured plan helps me stay consistent and make better decisions week to week.
Example of a typical training week I’m following:
Monday:
Rest day or very easy spin (45–60 min Z1) + short core session
Tuesday:
VO₂max session
e.g. 5×4 min @ 115–118% FTP — short, hard, high quality
Wednesday:
Endurance ride
1.5–2h Z2, steady and well-fueled
Thursday:
Threshold session
e.g. 3×12 → 3×15 → 2×20 around FTP,
Friday:
Rest or very easy recovery ride
Saturday:
Endurance or light sweet spot, depending on freshness
2–3h Z2 or controlled blocks
Sunday:
Long endurance
2.5–5h Z2, no intensity
Haven’t tried with Gemini but I did play around with this with ChatGPT last year.
I didn’t find it useful. It didn’t provide a sensical plan for me and did lots of really weird stuff. But I’ve also been training for years and can both recognize when it was dogshit, and be able to do my own programming without much difficulty.
If you were less familiar with training it might give you something that was better than nothing… but you also wouldn’t be able to identify when what it was telling you to do was nonsensical.
If you find it useful go for it, but don’t forget to bring your common sense.
I’m not sure how much it adds over just riding your bike as much as you can, with some intervals thrown in sometimes.
I don’t use it to develop my strength or cycling plan, but I do like asking it to do a recap of a workouts. It’s like checking in with a coach after a ride. I’ll show it all the stats from a given ride along with the nutrition I used on the ride and it gives me a nice summary of my performance combined with a break down of my fueling. I don’t do it after every ride, but use it as a check in on how I’m doing here and there.
Program Overview
No, sorry, I don’t have more to add. Just repeatedly doing high intensity stuff I couldn’t sleep well, couldn’t recover well, was stressed before training and saw it in HRV for a couple of days or more. Changing to threshold training, felt better, recovered better and HRV reflected that as well. It might be superfluous as I had enough indicators already. But it is nice to have some objective data as well to see that you are not making thins up.