TechRadar AI Week 2025
This article is part of TechRadar’s AI Week 2025. We cover the basics of AI and show you how to get the most out of the likes of ChatGPT, Gemini or Claude, along with in-depth features, news and the main talking points in the world of AI.
The next generation of Gemini could launch this week as rumors grow about the future of Google’s AI chatbot.
Gemini 1.5 surprised everyone, Gemini 2.0 felt like a reset, and Gemini 2.5 finally showed what Google’s AI can really do. On paper, it is powerful enough to compete with ChatGPT. In real life though, ChatGPT is still the chatbot people open first and the first thing that comes to mind when someone says ‘AI’.
Gemini 3 may be Google’s best chance yet to change that. To beat ChatGPT, it doesn’t just have to be smarter in a lab; it must be better to use every single day. Here are the five things Gemini 3 needs to improve if it wants to finally pass ChatGPT as the world’s number one AI chatbot.
1. Stop forgetting things
While Gemini 2.5 can technically handle giant conversations and long documents, it still loses track of what you’re talking about all too often. I’ve found that sometimes I’ll refer to something from earlier in the chat and it suddenly seems like it’s never heard it before.
It might not be a dealbreaker for everyone, but an AI chatbot’s memory is one of the most important features for the best experience, and I think Gemini is still lacking in that department.
Don’t get me wrong, ChatGPT isn’t perfect either, but it usually keeps the context more consistent and I’ve found its memory more reliable. Gemini 3 needs memory that actually feels human, and it should remember key details, understand ongoing tasks, and build on previous messages without forcing you to repeat yourself.
If Google gets this right, Gemini would finally feel like a proper assistant rather than a helpful stranger with short-term memory loss.
2. Respond faster but smarter
Response speed matters more than most people realize, especially when you want to have long conversations with your chatbot of choice.
Gemini 2.5 Flash introduced a think mode that shows its reasoning, and while I’ve been impressed with how smart it is, it can also be painfully slow.
ChatGPT 5 is now able to decide when to take longer to think depending on the prompt you give it, balancing speed and intelligence really well (most of the time).
If Gemini 3 can match the ability to determine how complex an AI model to use for each of your prompts and give users control, it would make Gemini instantly feel more responsive and more personal.
3. Get a better understanding of what you really mean
I don’t want this list to seem like ChatGPT is better than Gemini; in fact, I often find myself choosing Google’s AI chatbot over OpenAI’s. That said, there are so many areas where all AI experiences could be improved, including those I’ve mentioned and people like Anthropic’s Claude.
One of those areas is just better understanding in general; a more human-like understanding, one might say. There are moments when you ask Gemini a completely normal question and it returns something wildly random. It’s not that it can’t answer the question; it just guesses the wrong interpretation and confidently runs with it.
Gemini 3 must be better at reading intent. If a question can mean more than one thing, it should ask a quick follow-up instead of guessing. As I said, I’d like to see all AI chatbots improve on this, but Google was able to impress us with Gemini 3 and set the bar for its competitors to live up to.
4. Go beyond simple image understanding
Gemini can currently analyze your photos and short videos, but imagine if it could go beyond a superficial understanding and instead offer a real analysis of what it sees.
Imagine using Gemini at the gym and gaining a better understanding and analysis of your body shape, or true sous chef skills while cooking. Google has so much data at its fingertips; it would be nice to see the company capitalize on this and really transform the way AI can improve our lives.
5. Become the taskmaster we’ve been waiting for
Google keeps calling Gemini a true assistant, but right now it feels more like a very smart search box, and that can honestly be said of any chatbot, to be honest.
Gemini can find things and explain things, but when you ask it to do something right, it often hits a wall. It might show you a restaurant instead of booking it, it might draft an email instead of sending it, and if a step goes wrong, it gives up instead of adapting to solve your problem.
Gemini 3 should behave more like a personal assistant than a chatbot. This means performing tasks in multiple steps, handling changes on the fly, and correcting one’s own mistakes without starting from scratch. If it can plan a trip, organize your day, or manage a project from start to finish, people will finally see it as more than just another AI model.
Gemini 3 might launch this week, and if it does, I sure hope Google has some AI magic up its sleeve.
Follow TechRadar on Google News and add us as a preferred source to get our expert news, reviews and opinions in your feeds. Be sure to click the Follow button!
And of course you can too follow TechRadar on TikTok for news, reviews, video unboxings, and get regular updates from us on WhatsApp also.
The best video cameras
