Gemini felt dumb on December 5, 2025.

I put Gemini 3 and ChatGPT 5.1 through the same radiology and research tasks I use daily. ChatGPT walked me through detailed clinical reasoning, asked spot‑on follow‑up questions, and gave nuanced differentials that often hit the right diagnosis. Gemini’s answers felt shallow, skipping steps and missing key methodological details, even letting bad papers slip through. The contrast left me frustrated with Gemini’s oversimplification.

I started a coding project with Gemini 3.0, and after about 60 exchanges it completely stopped listening to my prompts. It kept bringing up old issues we’d already solved and suggested fixes that had nothing to do with the current problem, even though I spelled them out. On top of that, the Files system is a nightmare – I can’t delete anything, and Gemini renames files arbitrarily, spawning duplicate versions. The experience felt like being stuck on a broken 2.0 flash version.

I tried asking Gemini to colour a Manga panel, expecting it would refuse or mess up. To my surprise, it actually produced a coloured version. The result was decent enough to make me grin, and I felt a mix of relief and amusement seeing the tool handle something I thought it wouldn't. This unexpected success was a pleasant boost to my confidence in the AI.

I’ve been using Gemini Pro for a couple of weeks and was impressed—until today it totally stopped recognizing any images I send. A simple picture of a circle gets labeled as a ratchet, text becomes a treadmill. It used to be spot‑on just half an hour ago. Now it feels broken and really frustrating, especially since ChatGPT still gets the pictures right. I’m left wondering what went wrong.

I keep running into Gemini Live tacking on a prompting question to the end of every reply—no matter how I ask it to stop. It’s incredibly annoying and breaks my train of thought. I’ve tried explicitly telling it to let me drive the conversation, but it still inserts those unsolicited questions, which feels like a hidden system prompt forcing the behavior.

I spent hours testing Gemini 3 Pro Preview in AI Studio and hit a massive repetition bug. When the chat got long, the model completely ignored my latest prompt and copy‑pasted an earlier answer word‑for‑word, as if it’d forgotten the whole conversation. The attention seemed to drop early context, forcing me to keep re‑reminding it. The experience was frustrating and made any extended dialogue feel broken.

I tried to create images with Gemini's Nano Banana Pro just like yesterday, but today the option vanished. I'm over 18 and not using an organization account, yet the image generation feature is completely gone. It’s frustrating because it worked fine before, and I’m left clueless about why it disappeared.

I’ve been battling Gemini Pro’s endless hallucinations for three days. Every photo I upload in Thinking mode gets misidentified as a puzzle or some Russian book, and even the high‑resolution enhancement just spits back the original image. Switching between non‑Pro and Pro accounts didn’t help; the Pro version performed even worse, and it never seems to use the Gemini 3 model. I’ve tried both the Android app and the web app, but the issue persists, leaving me frustrated and unable to get any useful results.