Gemini felt dumb on October 19, 2025.

I keep hitting a wall with Gemini – every third question I ask gets a “I can’t help you with that” response, even though I’m not asking anything shady. I’m paying $125 a month for the Ultra tier, so the random censorship feels pointless and irritating. It’s been a frustrating back‑and‑forth that makes the tool feel unreliable.

I tried using Gemini for a Naruto roleplay and ran into a visual glitch that’s driving me crazy. The picture looks off, and no matter what I try, the AI won’t correct it. I don’t want to restart from scratch just to fix this annoying issue, so I’m looking for any workaround or tip to sort it out.

I tried to build my RPGs using Google AI Studio even though I don’t know any programming. At first it seemed promising, but as the game grew more complex the tool kept spitting out errors it couldn’t resolve. I felt stuck and frustrated watching simple bugs pile up, and the AI’s inability to fix them made the whole process feel like a dead end.

I tried asking OpenAI about a political topic and the response was heavily filtered, making the answer vague and unhelpful. It felt like the model had been deliberately throttled, leaving me frustrated that I couldn't get a clear answer. In contrast, Gemini handled the same query without any issue, which highlighted how restrictive OpenAI had become.

I asked Gemini to create an image, and instead it sent me a fake Imgur link like, “Here’s your download.” The URL led nowhere—twice. I never mentioned Imgur at all, yet it confidently claimed to have uploaded something that doesn’t exist. The hallucination felt eerie and frustrating, like the model was stuck in an old‑school API mindset. This broken behavior really undermined my trust in the tool.

I asked the model to generate a video about an airstrike over a city, and it churned out exactly that without any specific prompt from me. The result was unsettling and unexpected— I felt the tool crossed a line, producing dangerous content on its own and leaving me worried about its controls.

I generated an image with Nanobanana, saved it fine, but when I tried to re‑upload it for further editing the picture turned into a striped placeholder in the prompt bar. The whole creative flow stopped. I even tried Photoshop resaves and stripping metadata, but nothing helped. Google AI Studio shows the file correctly, so the problem seems specific to Nanobanana’s handling of its own output.

I tried using Deep Think after paying for Google Ultra, but I couldn’t upload my project’s code folder. The missing feature feels like a deal‑breaker, and I’m wondering if it’s been fixed or if it only works with GitHub. I’m hesitant to resubscribe until I know the upload issue is resolved.