Gemini users report simple questions now get template refusals
kueowirnzcd · reddit · 2026-11-02
A Reddit user reports that Gemini fails at basic factual questions that ChatGPT's free tier answers easily. Screenshots show Gemini's AI Mode replying with templated "consult official election authorities" refusals twice in a row, even after being pressed.
The author suspects this relates to reports that Google deprioritized Gemini in search, arguing it looks like the feature was effectively switched off, and asks whether others are seeing a similar capability drop-off.
More from Models
- Phil Chen: the biggest unsolved problem in LLM autoresearch is bypassing Anthropic safeguards — philhchen · 2026-11-02
- Opus 5.5 is good in exactly the ways DeepSeek is good, but not a deep thinker — teortaxesTex · 2026-11-02
- Quant-Arena benchmark: 5 of 9 LLM quant strategies lose money, none beat 17.1% baseline — jiqizhixin · 2026-11-02
- DeepSeek V4.1 Flash 1017 Scores 159 on Epoch ECI, Tied With GPT 5.5 — teortaxesTex · 2026-11-02
- Domingos: If OpenAI and Anthropic Had a Secret Sauce, Distillation Wouldn't Work — pmddomingos · 2026-11-02
- AI commentator laments frontier labs leak nothing, not even model configs — teortaxesTex · 2026-11-02