Gemini felt dumb on March 22, 2026.

I tried using Gemini and kept getting those stiff, scripted replies that just repeat my wording—like “you mentioned that you did X… you’re 100% right…”. The whole interaction felt overly robotic and irritating, making me wish the model sounded more natural instead of echoing my statements.

I tried using Gemini Vision to log data from screenshots, but the results were hit‑or‑miss. The upload process smushes the images into an ancient‑looking low‑res JPEG, so the model often misreads the numbers. I’m left frustrated because the tool can’t handle even basic sensor or spreadsheet tasks when the image quality is stripped away.

I’m beyond frustrated after the latest Gemini Ultra update. The model keeps losing context, ignoring documents I’ve uploaded, and even repeats old bugs because it forgets 80% of our chat history. Support was useless, and the output feels like early‑gen models again. Paying for Ultra feels wasted, and I’m ready to cancel all seven accounts if it doesn’t improve soon.

I was expecting a cutting‑edge response from their so‑called “SOTA” model, but it spat out utter nonsense. I didn’t add any system prompts or weird tricks, yet the output was garbled and useless. The experience left me furious and doubtful of their hype—feeling like I’d wasted time on a broken tool.