Gemini felt dumb on December 16, 2025.

I noticed my older Gemini chats have vanished—topics I discussed are gone, and even the search can’t find them. Google Activity still shows the prompts I entered, but the corresponding Gemini responses are missing or the link just loops back to the home page. It’s increasingly common, and the disappearing history feels really frustrating and unreliable.

I opened the web app expecting my Pro model to handle large prompts, but it felt like a tiny 7B‑parameter version with a 100k context limit. It couldn’t even follow the simple output format I specified, making the experience frustrating and unusable. Meanwhile, AI Studio ran smoothly, highlighting the inconsistency and leaving me disappointed despite my Pro subscription.

I’ve been using Gemini for a week and it’s getting noticeably worse. It’s slower, hangs more often, forgets context quickly, and ignores the data I explicitly give it. Issues that weren’t there at launch are now daily annoyances. I’ve even upgraded to Pro, but the problems persist. I’m wondering if I’m doing something wrong or if this is a known degradation, and I’d love any troubleshooting tips.

I’ve been playing with the Nano Banana Pro API for about a week and ran into two big annoyances. First, Google’s endpoint won’t give me exact image dimensions—I’m stuck with only resolution and aspect‑ratio settings, which feels like a missing basic feature compared to other models. Second, the API has slowed dramatically; a modest 1k image now takes 30 seconds or more. The slowdown and lack of precise sizing have been pretty frustrating.

I noticed Gemini stopped following the response style I set, and realized it couldn’t retrieve the “Instructions for Gemini” I’d written. Even after rewriting and confirming the instructions, the model still can’t access that info. It’s been puzzling and a bit frustrating, and I’m wondering if anyone else is seeing the same problem.

I’ve been stuck since yesterday with all five of my gems refusing to access any knowledge. No matter what I try, they act like a brand‑new chat each time, forgetting everything after just a few messages and not even knowing who they are. It’s ruined the seamless collaboration I had on my books, and I’m left frustrated and searching for a fix.

I tried to attach photos and documents in Gemini as I always have, but suddenly the option disappeared. Yesterday it worked fine, and now I’m stuck without a way to add files, which is really irritating and slows me down. The tool’s behavior feels broken, and I’m left guessing what went wrong.

I keep hitting “An internal error has occurred” every day for hours, making the service completely unusable. I’m stuck watching the same glitch repeat, and it feels like a massive waste of time. It’s frustrating to see a company with massive compute resources fail so badly, and it makes me doubt whether Google’s AI can ever be a reliable main app.

I tried to get Gemini to answer a prompting question after it generated an image, but it just spat out the same picture again instead of responding. When I followed up asking why, it replied with a random black‑and‑white photo that had nothing to do with our chat. The whole thing was oddly funny yet frustrating, leaving me shaking my head at its quirky misfire.

I was shocked when Gemini brought up a detail from a conversation I had two months ago, even after I deleted all my chats. I opened a fresh session, asked “What was the last thing we talked about?” and it somehow recalled that old exchange. It felt invasive and unreliable, making me question how the system handles “deleted” history.

I kept trying to get the frame‑to‑video feature to work, but it always crashes at 1‑2% no matter what image I feed it—whether a random plant photo or a cartoon dinosaur. The text‑to‑video prompt runs fine, so I’m stuck wondering if it’s a bug. The constant failures are pretty aggravating and halt my workflow.

I was trying to debug a massive codebase and after about ten prompts Gemini just dropped all the context I’d painstakingly pasted, acting like we were starting fresh. It’s the same model that used to remember thousands of lines for me, so losing the thread felt jarring and really slowed me down. I can’t use an editor‑integrated agent at work, so the web UI’s memory lapses are a huge hassle.

I kept trying to run Gemini Deep Research, but each time it spouted the “Something went wrong. Please try again.” error. Even though it counted toward my quota—showing I was “nearing my limit of 10 research reports”—the request never actually completed. I had to copy‑paste the exact prompt over and over, and this happened on almost every first attempt. The constant failures were irritating and made the tool feel unreliable.

I tried using Gemini 3 for tasks at work, but it constantly hallucinated, ignored my clear instructions, and even gave outright false answers. Every interaction felt like a gamble, leaving me frustrated and unable to rely on it. I ended up canceling the subscription and reverting to OpenAI’s plan because Gemini just wasn’t trustworthy.