I asked ChatGPT to write a 4-week strength training plan as a “Westside Barbell Trainer” to see what it would come up with - and its not terrible.
Like everyone else, I have been intrigued by Open AI’s ChatGPT since its release. I wanted to know if it could write me a strength training program, so I have been playing around with it to see what it can do. After several attempts it finally came up with something decent (the first few attempts were horrible). This article will discuss the prompts I used and assess the quality of the program ChatGPT provided.
Just as an aside: I wanted to try this because there are many people who will use ChatGPT for exercise prescription and health guidance. I want people to be informed of its strengths and limitations when it comes to exercise prescription. The best way to do that is to test it out and evaluate my results.
The first few times I asked for a training program I was not impressed with the “program”. It was often incomplete - very vague with repetitive information and no meaningful detail that you might get from a good personal trainer.
Here is a list of exercises it gave me from an earlier prompt I provided in February:
Sled pulls
Push press
Calf raises
Dumbbell curls
It should be clear that there is absolutely zero cogent thread to be pulled from this. The exercise selection is horrible. No information on sets or reps. No rhyme or reason. Yikes.
So after several instances with results like this, I came back with a new approach. I knew ChatGPT needed a framework to base the training on, a website that Open AI’s LLM would have crawled back when it was training the model. So I decided to go with Westside Barbell’s training methodology. Westside BB has been around since the 70’s and is known for producing a large number of very strong people. Most importantly WSBB publishes a lot of their content for free which would give the LLM access to the knowledge base.
Westside uses what is called the conjugate method - combining both Soviet and Bulgarian influences. The training is mainly centered around the big three lifts, Bench, Squat, Deadlift. The traditional conjugate system is 4 days / week (Mon, Wed, Fri, Sat) where athletes trained maximal effort (1-3 rep maxes for 3-5 sets) or dynamic effort (performing each rep as fast as possible) on a given day. Dynamic work is done with chains or resistance bands. Either mode is split into upper or lower body days, and supplemented with accessory lifts. Accessory lifts use the repeated effort method where most sets go to failure or close to failure. So if we take all of that together one week would look something like this:
Monday Day 1: Maximal effort lower body + Repeated effort accessory lifts
Rest day
Wednesday Day 2: Maximal effort upper body + Repeated effort accessory lifts
Rest Day
Friday Day 3: Dynamic effort lower body + Repeated effort accessory lifts
Saturday Day 4: Dynamic effort upper body + Repeated effort accessory lifts
This is, to the best of my knowledge, a rough outline of what you might get from a real (human) WSBB coach. As always the devil is in the details, so what exactly one might do depends on the athlete (their competencies / weaknesses etc.) So the question is, can ChatGPT provide something that feels like WSBB?
There is only one way to find out.
Here is the full prompt I gave ChatGPT to write a strength program as a WSBB trainer.
As you can see ChatGPT responds with a decent amount of detail. I would have liked a defined series of exercises for the accessory lifts rather than a list to choose from. Furthermore, I suspect that WSBB would program more than 2 accessory lifts… So not quite on the mark there.
However, I will say that the general structure is pretty close to the conjugate method. One issue I see with the general outline of the program is that there is no detail about what days of the week to perform the workouts. WSBB recommends one day of rest between max effort days and the following workout. As you may notice, there is no information about the days of the week these workouts should be performed.
Okay so there are some issues, but this is par for the course in my experience. And to be honest, someone with not exercise background may have no issue with this program. But that is the point of this article.
Here is the rest of the general structure for week 2-4.
I then asked ChatGPT to put the program into a table format so it was easier to follow. Here are those results:
Week 1 of ChatGPT WSBB Strength Training Program
Week 2:
Week 3:
Week 4:
As you can see its generally providing the right structure based on the Westside Conjugate system. But the precise choice of lifts, particularly the accessory lifts are not great. ChatGPT seems to have a weird obsession with calf raises… Which in my opinion calf raises are a complete waste of time… If you want to workout your calves, just go for a run.
All that said, I am actually going to give this program a try (minus calf raises) and document it along the way.
Even if it doesn’t produce results, I think it will be a fun little experiment. Follow along if you find this interesting!
Here is a quick peak at day 1 of the program
