Why I Switched from ChatGPT to Gemini After 3 Years (2026)

I subscribed to ChatGPT roughly 3 months after it launched, somewhere around early 2023. Used it pretty much every day, kept the subscription going through all the model updates. Got a lot of value out of it over those three years.

I switched from ChatGPT to Gemini in December 2025 and haven’t gone back.

Not saying ChatGPT is bad — I wouldn’t have stayed for three years if it was. This is more just an account of why I eventually felt the switch was worth making, and why it took as long as it did.

Why ChatGPT Started Feeling Off

A few things had been sitting in the back of my mind for a while before I actually did anything about it.

ChatGPT felt like it had grown complicated. The product expanded in a lot of directions and navigating it started to feel less clean than it used to. I’m not sure if I just changed how I used it or if the product genuinely shifted, probably a bit of both. Either way it stopped feeling as simple as it once did.

I also noticed I’d stopped getting excited about updates. Early on each new model felt like a real step forward. By 2025 the updates were still good but my reaction had become pretty neutral. That kind of slow drift is easy to miss until you look back and realise it’s been happening for a while.

None of this was enough on its own to make me switch. I’d built habits around ChatGPT that worked and I don’t tend to jump to new tools for the sake of it.

The Context Window Gap: GPT-5 vs Gemini 3 Pro

I was messing around with using an LLM as a solo D&D dungeon master. It’s something I’d been interested in trying for a while. Sessions get long and context-heavy pretty quickly, and I kept running into coherence issues — the model would start losing track of things that happened earlier in the session.

At the time GPT-5 had a 256K token context window in the ChatGPT interface. Gemini 3 Pro sits at 1 million tokens. That’s roughly 4x the context, which in practice made a noticeable difference for something like a long D&D session where you’re carrying a lot of state. I tried it with Gemini and it held up better. Not perfect, but better.

That was the point where the question shifted from “is there a reason to switch” to “is there actually a reason to stay.”

Gemini 3 had just launched in November 2025, so the timing worked out. I’d been using it for a couple of weeks on the free tier before I made the call to actually subscribe and drop ChatGPT.

Why Not Claude for General Use?

Worth bringing up since I use Claude heavily for work through Claude Code.

Honestly it’s mostly a psychological thing. I’m on a Pro plan and I use Claude a lot for actual development work. When I open a chat to talk through something casually, there’s a voice in the back of my head that goes “this is eating into my rates.” That feeling makes the experience worse even if it probably isn’t accurate.

Gemini on a subscription doesn’t carry that association for me. So I’ve ended up with a natural split — Gemini for general use, Claude for work. It’s working fine. Whether Claude would actually be good for casual use on its own terms I genuinely don’t know, I haven’t given it a proper go in that context.

Was Switching from ChatGPT to Gemini Worth It?

A few months in and the switch has stuck. The context window is still the main practical reason and the general quality has been good enough that I’m not missing anything day to day.

If you’re happy with ChatGPT there’s probably no urgent reason to change. Different tools end up fitting different people for different reasons and a lot of that depends on what you’re actually using it for.

For me the context window was the concrete thing that made it feel worth switching. Everything else was just the slow build-up that meant I was ready to make the call when it came.

I cover where Gemini fits in my broader AI setup — alongside Claude Code and OpenRouter — in my LLM tools breakdown if that’s useful context.