Ryan Haines / Android Authority
I’ve lost almost 9 pounds (4 kg) so far, my upper abs are starting to show, and I just generally feel a lot better, all thanks to Gemini. The chatbot has helped me come up with a detailed training as well as a diet plan and has guided me through all the issues I’ve run into during my fitness journey that kicked off back in September.
I wrote about my experience using Gemini as a personal fitness coach in November, and it was mostly positive up until that point. However, things quickly took a turn for the worse. The chatbot went from a proper fitness coach I relied on daily to nothing more than a distraction that caused more harm than good.
Would you trust an AI chatbot to be your primary fitness and diet coach?
3 votes
Yes, it’s better and cheaper than a human.
33%
Absolutely not. I don’t trust robots.
0%
I’m on the fence about it.
67%
A case of digital amnesia
Edgar Cervantes / Android Authority
I wanted to keep all the data about my fitness journey in one place, so I had a single chat open in Gemini that I used daily. The problem is that the longer the conversation went on for, the more issues I encountered.
Gemini kept forgetting things. It forgot my training plan altogether, so when I asked it to show me the workout scheduled for a specific day, it just started hallucinating and gave me a random one that wasn’t aligned with my goals or the fitness equipment I have available.
I tried telling it that the workout suggested wasn’t the one in my original training plan and that it should try to pull up the right one. It tried, but failed. Over and over again. The only way to find my complete training plan was to scroll back through months of conversations and try to locate it manually, which isn’t really how I want to spend my afternoons.
Don’t want to miss the best from Android Authority?
Aside from my workout, Gemini also completely forgot my diet plan. It was well thought out and detailed, containing the exact calories for each day along with the protein count. When asked to show it to me, the chatbot just made one up. What’s worse, it even forgot that I’m a vegetarian and listed a meal plan full of chicken and other meats.
Prompting it to try and pull up the correct info didn’t get me anywhere. At one point, Gemini started asking me about my diet, including the caloric and protein count I planned to consume, so it could log it in its memory. I mean, if I remembered the whole thing by heart, I wouldn’t need Gemini in the first place.
The AI’s forgetfulness didn’t end there. Gemini didn’t know how many weeks I’ve been training for and pulled up the wrong body measurements when I asked it. Honestly, the whole thing turned out to be a disaster, and I was willing to call it quits.
It was just a matter of time
Kaitlyn Cimino / Android Authority
There is a reason for this behavior, and it is a common issue among chatbots — so it isn’t isolated to just Gemini. In simple terms, every AI chatbot has what’s called a context window, which acts as its short-term memory. This window has a limit on the number of tokens it can hold at any given moment. A token can be a whole word, part of a word, or even a punctuation mark. On average, 1,000 tokens equals around 700 words.
As soon as you reach the token limit of the context window, old information is discarded to make room for new input. This means Gemini forgets earlier details once they leave its context window. Unfortunately, instead of admitting it doesn’t know or have access to the requested info, it often starts hallucinating — making things up to fill the gaps.
The main problem with tokens is visibility: as a Gemini user, I can’t see how many I’ve used. I often have to wait until I run into issues with forgetfulness to realize I’ve hit the limit.
The ‘lost in the middle’ effect contributes to Gemini’s forgetfulness.
There are other issues that contribute to Gemini’s forgetfulness, such as the “lost in the middle” effect. This refers to the tendency of AI models to focus more on information at the very beginning and end of a conversation, often losing track of details buried in the middle. This becomes especially apparent during long conversations with a lot of complex information.
There’s a solution (kind of)
Joe Maring / Android Authority
The solution is far from ideal, but at least it’s a path forward. I had to start a new chat and leave the old one — along with months of valuable data — behind. However, I’m changing my strategy for this new thread.
Instead of going back through the long process of having the AI help me build a plan from scratch, I mined my old chat for all the vital details and fed them into the new one to get Gemini up to speed. But I realized there’s a much smarter way to do this than just pasting text: storing everything in a Google Doc.
I’m now keeping all my essential data in a dedicated file, including:
My exact training and diet plans.
Every measurement I’ve taken throughout my fitness journey.
Progress pictures and historical data.
My long-term goals and general vision.
This way, if I hit a token limit or Gemini randomly starts forgetting things, I can simply reshare the link to the doc to refresh its memory. It also provides me with the benefit of having a clean and organized view of my own data.
If I ever hit another roadblock where the AI becomes a source of frustration, I can start fresh in a matter of seconds by pointing a new chat toward that Doc. Honestly, I should have done this from the very beginning, but I was so eager to get started and didn’t realize I would hit these limits so quickly. Lesson learned.
Thank you for being part of our community. Read our Comment Policy before posting.
