For clarity, comparing TR AI released in January 2026 versus the non-AI TR platform prior to 2026.
This is a completely unofficial poll!
my vote for “somewhat worse” is mainly due to the outdoor workout bug, as I actually can’t use the aiFTP at this time and have to set FTP manually. Which defeats the purpose of it.
What’s the outdoor workout bug? I’ve been doing outdoor workouts and haven’t noticed an issue
some users doing outdoor workouts are seeing an aiFTP drop after successfully completing their workouts.
I think the concept is better. The capability of taking in and utilizing unstructured rides is huge as far as I’m concerned and it has worked kinda sorta OK for me when I was on a plan earlier this year. There’s been a lot of changes, hiccups, and bugs. It’s pretty complicated and knowing the proper way to optimally use the features is not clear, and explanations and instructions have not been consistent. I expect to jump back in Nov or Dec and look forward to see how it’s evolved by then.
I’m kind of the same, hoping to return in September/October and find things have settled down. For now, the weather is too good to do anything except enjoy being out on the bike.
While there are definitely many small and large improvements, the key feature, providing me with the right workouts to get faster, does not work for me since at least four weeks.
It looks like something in my history brought the recommendation engine off-track.
That’s what happened to me this past winter. The weather was great so rather than toil in the pain cave, I rode outside. I don’t bother with structured workouts outdoors so it was great that TR utilized the data from my unstructured rides. That is awesome, but what would be awesomer is if would provide some guidance for unstructured rides beyond the simplistic red-light (yellow light) green-light feature.
What I would like (and I know it’s probably asking a lot) is that unstructured outdoor rides would factor better into FTP estimations …Like if I do 20 mins @ 350W on a climb for instance, then it should be obvious that my FTP isn’t 275W …Would save a lot of faffing around when I return to indoor training.
It more or less that for me. I did few workouts and mainly unstructured rides. I made no changes to the workouts in the plan and just let them expire, It continued to update my “AI FTP” the day after each ride. Now, that wasn’t the same thing as doing an FTP test, or even a de-facto FTP test like you did, but my predicted FTP and scheduled workouts updated after each ride. I expect that when I start training again, it’ll take all my data up to that point and create an appropriate plan.
FTP Detection will work this way when this fix rolls out. That issue is being addressed by the team right now. 
My general sense is that it’s somewhat worse. I believe it over weights TR workouts. On Saturday I skipped my planned 30 minute workout and did a 75 minute MTB ride. Hard. Double the planned TSS and PR’d on the route. But I “skipped” my planned ride. I assume it took that as a signal that I’l skip future workouts, so my FTP declined. Maybe it’s just me but this seems common.
Fantastic! Thanks Eddie.
Just want to say it’s really nice to see people voting who aren’t frequent posters/“likers”. It’s nice to know you’re all out there, even if you’re not active in the conversation. Sometimes I feel like there’s maybe 20 of us here and it’s nice to see other opinions!
I feel like an outlier, but so far AI detection has been working great for me. It may be that most of my training is on a trainer.
Notes:
Positives:
Negatives:
Will be interesting to see how the poll shakes out as more and more people respond, but right now it is 40-50% of responders say that TR is slightly worse or significantly worse. That isn’t just squeaky wheels. That is feedback that would raise action where I work, and likely illicit some significant changes both behind the scenes and perhaps to our product as well.
Those are people who do not understand how Ai works or are upset that not following the plan does not make them faster.
And people who are happy are far more likely to make excuses for why everything is just fine.
I don’t care much about detecting or predicting my FTP using “AI,” but I find the new UX to still be very frustrating.
This is the first I’ve complained about it because I’ve wanted to give it some time to improve (and for me to adapt), but it’s honestly just a pain. I feel as though I’m forced to wade through a feature set that I didn’t ask for before I can complete core tasks in the app.
I’ve been a subscriber since 2018 and will not be renewing this fall.
To be fair, people who are unhappy are pretty good at excuse-making as well. I have seen a fair few haters come through who are unwilling to have their training evaluated at all or when its pointed out to them that their methodology is deeply flawed they either dig in to their misunderstanding of the tool or disappear entirely. That is not to say that’s everybody mind you, I know there are plenty of people who are legitimately struggling with the AI and I hope TR gets a handle on the consistency issue across athletes. But the people who come on here - and they have been many in number since the beta - who are barely following any plan at all, doing what they want, ignoring the plan, skipping workouts, and complaining loudly and angrily that it doesn’t work, well they do no favors for the riders who are doing everything right and still not seeing results.