Gemini wishlist: These are the features I want to see

I’ve been using Gemini Advanced for a while now, and for the most part, I’m really enjoying it. It significantly speeds up my workflow, whether I need brainstorming help for work or I’m diving deep into a topic I’m currently obsessed with.

But while I find it tremendously useful, there’s definitely room for improvement. There are features I find myself missing when using Gemini — capabilities that I believe would make this already great AI chatbot even better. I’ve compiled a list of five key areas for enhancement. While some might seem ambitious for now, I’m optimistic we’ll see them implemented eventually.

I rely on numerous apps daily, and constantly switching between them disrupts my focus and efficiency. Ideally, I envision Gemini acting as a central hub, allowing me to interact with and retrieve information from my most-used applications without leaving the Gemini interface.

For example, an integration with a task management app like Todoist would be incredibly useful. I could simply ask Gemini to display my tasks for today or the whole of next week. Adding new tasks or rescheduling overdue ones directly through Gemini would be a significant time-saver.

Similarly, I’d love integration with my money management app. Imagine asking Gemini to pull up my budget, report my spending on dining out this month, or check the balance in my “Travel” savings category.

While I still see value in dedicated apps for specific functions like task and money management, having a central command center like Gemini that can access data or control app functions via prompts would streamline many daily activities. Gemini already integrates with Google’s own ecosystem (like Gmail and Maps), and these integrations work reasonably well most of the time. However, the potential for third-party integrations feels largely untapped, at least officially.

Gemini’s image generation, powered by Imagen 3, is impressive. I’ve experimented with it quite a bit and am generally impressed by its capabilities. Sure, some results can look overly processed or occasionally turn out a bit strange, but it’s a powerful tool, especially considering its relative newness.

However, the editing capabilities leave much to be desired. For instance, if I ask Gemini to create an image of a cat wearing a funny orange hat, it typically delivers. But if I then follow up with a prompt to change a specific detail — say, changing the hat color to blue — Gemini often regenerates the entire image.

The cat might now have a blue hat, but its pose, background, and other elements often change, too. Ideally, I should be able to lock the existing image and prompt Gemini to modify only specific elements—like altering someone’s eye color, adding items to a desk, or removing unwanted objects — without affecting the rest of the composition.

I recently moved into a brand-new apartment, and furnishing it was a challenging process. I ended up creating a 3D model of my apartment using specialized software, decorating it with available furniture assets and some specific 3D models I managed to download and import.

The final visualization was amazing, but achieving it was very time-consuming. I envision a future where Gemini could dramatically speed up this process. For instance, I could upload a floor plan with exact wall measurements, and Gemini would generate the 3D model. It could then automatically furnish the space based on detailed prompts. I could specify my preference for bright furniture and a Scandinavian interior design style, then request specific adjustments, like making the sofa larger and L-shaped.

A major hurdle I faced was finding 3D models for specific furniture items I had purchased, making my manual 3D model less realistic. While models for popular IKEA items are relatively easy to find, less common pieces are often unavailable. It would be incredibly useful if Gemini could generate a 3D model from an image of a specific item and then place it within the 3D apartment model it created. With these capabilities, I could visualize my fully furnished apartment perhaps ten times faster than doing it all manually.

Gemini and Gemini Live boast support for over 40 languages, which is impressive on paper. However, the quality isn’t uniform across all supported languages. There’s a noticeable difference between using Gemini in English versus Slovenian, my native language.

This disparity is especially pronounced with Gemini Live. When I interact with Gemini Live in English, the conversation feels remarkably natural, almost like talking to a person. Switching to Slovenian, however, degrades the experience significantly. The accent is unnatural, and the voice sounds robotic, making it unpleasant to use. Consequently, I stick to English for voice interactions.

The text-based chatbot fares better in my native language than Gemini Live, but it’s not without its quirks. Occasionally, even if I pose a question in Slovenian, the chatbot inexplicably responds in English.

I can’t speak for other languages, but I suspect widely spoken ones like Spanish or French likely perform better than Slovenian. I’d be interested to hear about others’ experiences in the comments.

A fundamental challenge with Gemini — and most current AI chatbots — is their propensity for errors. They can “hallucinate,” meaning they generate plausible-sounding but incorrect or fabricated information. While not constant, these inaccuracies occur often enough to undermine trust.

Trust is paramount. Knowing from experience that Gemini might provide misleading or outright false information makes me skeptical of every response. This skepticism is particularly problematic when seeking information on critical topics like legal, medical, or financial matters, as verifying the accuracy is difficult and the stakes are high. For this reason, relying solely on chatbot advice for such sensitive areas is currently not smart, though I hope this changes soon.

I look forward to a future where hallucinations are significantly reduced, if not eliminated, allowing users to trust the vast majority of Gemini’s output. Imagine receiving a reliable preliminary medical assessment based on symptoms or accurate legal pointers that save you an initial consultation fee. That requires a level of reliability we haven’t reached yet. I understand this will take time as Google continues to refine its AI technology, but it’s a crucial step. Make it happen, Google.

Now it’s your turn — let me know in the comments which features you’d like to see added or improved in your favorite chatbot, whether it’s Gemini, ChatGPT, or another AI assistant.