Gemini felt dumb on March 9, 2026.

I’ve been playing with Gemini lately and was blown away by how well it now cross‑references old chats. I can mention a topic from a year ago and it instantly pulls the right context, which feels like a massive quality‑of‑life boost. It even grabs relevant details—like “green chairs”—from past conversations to explain things clearly, making the tool feel far more useful and intuitive.

I’ve been testing Gemini 3.1 Pro and every time I upload a photo it spouts huge hallucinations—mistaking a person’s face for a dog and stubbornly insisting it’s a dog even when I correct it. It started last night and has kept happening across several chats. The image analysis feels like it’s regressed to rock bottom, making me hesitant to trust it for anything simple now.

I tried using Gemini to get news updates, but it completely missed the mark—its responses were nonsensical and even seemed to “play along” with my prompts in a misleading way. The tool’s behavior felt dangerous, as it could easily spread misinformation, and I felt frustrated and alarmed by how useless it was for a simple task like reading the news.

I tried the Gemini 3.1 Flash Lite preview and was left genuinely disappointed. The responses were shallow, often missing the point, and the model struggled with even basic prompts. It felt like the tool couldn't keep up with my requests, making the whole experience frustrating and far from what I'd expect from a Google model.