Gemini starts refusing fitness and nutrition advice after working normally
Ignaciovr · reddit · 2026-07-23
A Reddit user says Gemini stopped helping with training and nutrition planning after previously working well with imported ChatGPT memory, uploaded files, and a custom GEM prompt.
According to the post, the model now drafts an answer and then replaces it with a safety-style refusal like: “This conversation is not my thing. I can’t help with anything that may not be safe or appropriate.”
The user asks what can be done to restore the old behavior, suggesting a change in Gemini’s guardrails or response policy on fitness and diet-related advice.
More from Apps
- One-prompt 3D muscle anatomy site built on GPT-6 via WorkBuddy — lxfater · 2026-09-11
- HR worker builds AI agent workflow for recruiting, saves 5-6 hours a week — scorpi1011 · 2026-09-11
- You Don't Need the Best Model: Blogger Uses Gemini in Gmail to Sort Out His 401k — 0xsachi · 2026-09-11
- Bolter launches as one of the first truly collaborative AI apps — Dan_Jeffries1 · 2026-09-11
- OmniGet: open-source app downloads and transcribes content from 1,800+ sites — tom_doerr · 2026-09-11
- Replit users are having fun building 3D childhood bedrooms with Astra — amasad · 2026-09-11