Gemini felt dumb on October 28, 2025.

I was working in AI Studio when Gemini suddenly went haywire. It started spitting out nonsensical answers and completely misunderstood the prompts I gave it. I tried to steer it back, but each response felt more off‑track, making the whole session a frustrating chase. The experience left me annoyed and skeptical about relying on it for any serious task.

I built a tiny gem to fetch The Mentalist episode plots by feeding it an “S05E14” tag, and at first it nailed the details. After a week, every query started spitting out completely made‑up synopses, even though it still got the episode titles right. I’d have to retry two or three times before it finally gave the correct plot, which was frustrating because I rely on it for quick, casual look‑ups.

I tried using Gemini Live on my phone and kept getting weird follow‑up prompts at the end of every answer, like “Would you like to learn more?” It was annoying and broke my flow, especially since the desktop version doesn’t do this and other bots behave normally. I even asked it to stop, but after a few turns it fell back to the unsolicited questions, making the experience frustrating.

I tried to use AI Studio, but the scroll kept jumping to random sections. I couldn’t edit the latest messages, read the conversation, or jump to a specific point. The broken scrolling made the tool unusable, leaving me stuck and frustrated.

I’ve noticed the AI Studio stops “thinking” after a few questions—it just picks options without any deliberation. It feels like the tool loses its depth early on, and I’m wondering if paying would bring back that thoughtful behavior. The experience has been disappointing and confusing.

I set up custom instructions for my gems and they work fine in AI Studio, but once I run them in Gemini they go off the rails. The responses become garbled, repetitive, and almost nonsensical, like “Okay… that's… new… Words… not… working…”. It used to be fine a week ago, and now the output just degrades quickly, which is really frustrating.

I ran a side‑by‑side test of ChatGPT, Claude, and Gemini to see if they could actually back up claims with real citations. Gemini completely tanked – zero correct references out of everything we threw at it. The experience was alarming; I felt the tool was unsafe for any academic work and had to scrap it entirely, resorting to manual checks or dedicated citation services.

I tried to use Google AI Studio but the scrolling bar was broken, making it almost impossible to read any replies. When I attempted to edit a response, the characters vanished the moment I clicked edit. The whole experience felt sluggish and careless, leaving me irritated and questioning why the team let such basic UI problems slip through.

I was relying on Gemini to craft my study plan, setting up details and trusting it to keep everything organized for my upcoming exam. Suddenly the whole chat vanished—every file, every step erased, leaving me back at square one. The loss was stressful and disrupted my prep, and I’m left scrambling to rebuild everything before the test.

I was excited to use Genimi Storybook for both teen and adult stories and got decent results at first. But when I tried to extend a story into a “Part 2,” the system threw errors and stopped me dead in my tracks. It was frustrating because the initial output was good, yet I couldn’t continue the narrative.