Testing Gemini Deep Research on a GEO investigation: it missed key points and got things wrong
dejanseo · x · 2026-11-14
SEO/GEO practitioner dejanseo asked Gemini + Deep Research to produce an investigative-style report on Generative Engine Optimization (GEO). His verdict: mediocre — it missed many important aspects and made claims he disagrees with, such as the idea that changing content does change model preferences. He published the output for others to judge. The takeaway: Deep Research-style products still need domain-expert review, with clear gaps in factual accuracy and judgment on niche topics.
More from Models
- Uganda team ships 350M-param Luganda chat model, but compute is exhausted — weskambale · 2026-11-14
- Unsloth ships Dynamic v3.1 quantizations for Qwen3.8-27B local deployment — slacka123 · 2026-11-14
- Turing Award winner Patterson praises GrokBot: better and faster than humans — if that's not AGI, what is? — davidpattersonx · 2026-11-14
- User reports Opus 5.5-designed logic analyzer chip works flawlessly after fab — IgorCarron · 2026-11-14
- Baidu's Ernie 5.5-Flash preview tops GameFeel benchmark as strongest open model at 280B — teortaxesTex · 2026-11-14
- Japanese inference provider ai& day-one launches Kimi K3.1 Pro and Muse Spark Open — DavidBennett__ · 2026-11-14