Gemini felt dumb on February 17, 2026.

I tried using Gemini 3 Pro’s web app and was shocked by how badly it performed. Every question I posed got garbled answers, with the model hallucinating basic facts and even twisting my original prompts. The constant mis‑interpretations and gaslighting made it feel infuriatingly useless, turning a once‑helpful tool into something I couldn’t rely on at all.

I subscribed to Gemini Pro expecting 100 daily posts, but now I only get 5‑10 before I'm told to wait 3‑4 hours. The reset used to be simple at 1 am, but now it’s a scattered, piecemeal limit that interrupts my workflow. It feels like the same restrictive throttling Claude uses, making the tool unreliable and almost useless for continuous work.

I tried using Gemini’s web/app to pull info from a URL, expecting it to actually read the page like AI Studio does. Instead it acted like a lazy Google snippet search, guessing content and giving wildly off answers. The mismatch was irritating, and I’m left wondering if there’s a hidden trick or if this limitation is just built‑in.

I keep hitting Gemini for anything that needs up‑to‑date info and it just hallucinates, even with the PRO toggle on. When I ask it to “search the web,” it rarely actually pulls anything, and the few sources it does give are sketchy compared to ChatGPT or Claude. As a Google product, the lack of a real Web Search button feels like a huge letdown and makes the tool frustrating to rely on.

I asked Gemini 3 Pro for a precise, sourced risk estimate of schizophrenia for a teen who smokes cannabis and has two uncles with the disorder. It confidently gave a 75% cumulative risk, citing studies, but a specialist psychiatrist told me that figure is far too high since uncles aren’t first‑degree relatives. As a clinician who paid for the service, I felt uneasy and misled, worried the tool might steer my decisions down a dangerous path.

I tried using Gemini Flash and was pleasantly surprised—it felt crisp, speedy, and consistently delivered exactly what I asked. When I switched to Gemini Pro, though, the experience turned sluggish; it dragged on, often left me with unfinished or incorrect output, and practically stalled my workflow. The contrast left me frustrated and wishing the tool behaved more like the Flash version.

I kept trying different models and each one threw the same cryptic error screenshot. I wasn’t even using any model earlier today, so this was my first attempt and it failed instantly. The tool’s behavior was baffling and frustrating, leaving me stuck without any progress.

I compared ChatGPT and Gemini and found Gemini less formal but more playful, with lighter moderation. ChatGPT feels heavily censored and lecture‑like, though its reasoning, web search, and overall intelligence impressed me. The mix of professionalism and overwhelming tone left me with a mixed, somewhat frustrating impression.