Long Story short, going into the last week of my block, the AI ftp prediction had been going from 275-304 with 7 days and 3 workouts to go. After doing 2 of the 3 workouts, it’s reduced my prediction by 10watts. I did the workouts outdoors which added about 20 minutes of endurance in order to ride to my preferred interval location. I feel like I did a fairly good job of hitting my interval targets. The most recent workout was two 2 minute blocks of sweet spot and I felt like my heart rate was lower than expected for those efforts. The AI prediction tool straight up sucks. It’s completely deflating to feel like you’re hitting all your targets and continually watch your “payday” shrivel up and die. I will make my calendar public and I give anybody permission to comment and share results to this forum.
I’ll take a closer look, but it’s first important to know that there is a bug that we’re currently working on squashing that causes AI FTP Predictions to start high and then fade away as you approach your FTP Detection date when you’re doing your workouts outside.
It sounds like this is what you’re experiencing.
We’ve got our best people on the case. Sorry for the trouble!
Hey @lrh8197 
Just wanted to add to @eddie post that while this but may be affecting you due to the outdoor workouts, there are other things that may also be naturally driving your FTP Prediction down, even though your FTP Detection turned out higher than the last, which I think should not go unnoticed (274>> 283) !
Things potentially driving your FTP Prediction down over time:
- Looking at your training plan, you have 3 scheduled interval workouts per week but only hitting 2.
- While it’s OK to add unstructured rides, these 1) don’t have the structure that builds FTP gains, 2) can cause a negative effect by building unwanted fatigue>> lower FTP gains. Having a good balance between recovery and training stimulus is key.
- Manual changes to the recommended TR Workouts can impact FTP gains. For example, April 5th workout was manually picked and April 26th workout was an Alternate that got rated Very Hard. While this is OK to do as well, it’s another factor that can shift the prediction.
- This is more something to keep in mind… When you complete a workout, it’s not that one workout that changes the prediction, is the recalculation of the entire 28 days (what has been completed + what is expected to be completed), so naturally, with the changes above, your prediction will change over time.
I also noticed that you started your plan March 24, 2026 and just got off the Base Phase and into the Build Phase. This is a good side thing to keep in mind as well as we see the most FTP gains during the Build Phase, so moving forward, (if you would like FTP gains), I would focus on consistency with keeping with the plan, or even changing the plan to just 2 intervals per week + 1 Solo Ride so you can keep enjoying your unstructured riding while balancing training vs. recovery.
Hi Caro. Thanks for taking the time to look at my calendar and providing some perspectives.
I have been using trainer road for several year and have always had good results. I like the new AI for its ability to pick and progressively overload workouts, and I do think it’s a big improvement to the way that training plans are organized.
My biggest issue is definitely the FTPP. Historically, I have made the biggest gains with higher volume and 2 days of intensity. This year I decided to let the AI ftp take the wheel completely because I was curious about the new AI and interested by the predicted 20-30w gain for a single block.
That prediction was still predicting 300+ watts at the beginning of the week leading up to detection. My actual AI ftp detection was close to 10 watts lower than what the FTPP settled on.
I feel like the overall messaging of the FTPP product is that rigidly complying to the plan will deliver incredible results, but then the rug gets pulled out at the end of the block.
Now I’m looking back at the block with some amount of regret at the rides that I COULD have done out of enjoyment but passed on because I thought I was supporting a magical 30w ftp jump 
.
Don’t get my wrong, 10 watts is fine but in the past I’ve had similar or greater results at the beginning of a block just from riding endurance.
I’m still a fan of trainer road overall, so don’t think that I’m on the edge of rage-quitting or anything, but my genuine feedback as a long time user is that the FTPP is a bad enough feature that it will turn off a lot of users to the overall product. It feels gimmicky.
Does that include unstructured rides with heart rate data only? Trying to determine if riding outside for fun after a longer winter is as detrimental for my fitness as AI FTP thinks it is.
I’m not sure I fully understand your question. 
I wouldn’t worry about things too much. Ride outside for fun whenever you’d like. Send us as much data as you have and don’t worry about the rest.
Your fitness will be clear when you train with power data. 
Sorry, I understand the confusion since you obviously don’t know my specific situation. Basically I was training indoors pretty consistently leading up to a gravel race on 4/12. Decided at that point I could use more time riding outside. Haven’t been back on the trainer since (the weather is nice and I just got a bike fit on my race bike), but meanwhile AI FTP thinks I’m hemorrhaging fitness at the moment. But you’re right that I’ll know better once I actually get back on the trainer and do a workout. Feel free to look at my calendar if you want (I have a pretty thick skin). I believe it is set to public.
I probably should just put a power meter on my gravel bike
If you’re riding outside for the time being, don’t worry too much about your FTP. Just enjoy the ride.
If you don’t have a power meter on your bike, you should use HR or RPE for pacing rides and races anyway.
I’ve turned off the AIFTP Prediction, I don’t want to be a slave to it, my plan is two interval sessions mid-week along with the easier rides prescribed by plan builder , although if the weather is nice I do plan on doing a few 90 minute post work rides on Wednesday evenings, and then my weekends will be riding for fun and audaxing at the weekends.
One of my biggests regrets about my racing years, is the amount of days out with friends I passed up on chasing FTP and time trial times, those times will never come back and I would happily trade a few watts now for a great day out on the bike with friends.
Definitely sounds like what’s happening to one of my friends who does TR workouts outside. His 200>>221W forecast became 200>>199W by the end 
I suspect I’m getting this inside too, my last workout (before detection date) was predicted to be “Hard” (leaning towards “Very Hard”) and after finding it “Hard” my forecast dropped by 2W. Maybe it’s because I hit a new PB for max heart rate though 
Hey @lrh8197, this is great and genuine feedback! I feel you.
I will make sure to share it with the team.