Google is rolling out a wave of enhancements to its Gemini AI app

The latest update introduces an upgraded research assistant, deeper personalisation, and expanded connectivity with Google apps

Google is rolling out a wave of enhancements to its Gemini AI app, aiming to make it more intuitive, powerful, and deeply integrated into users' lives. This update brings an upgraded research assistant, deeper personalisation, and expanded connectivity with Google apps — several of which are being made more widely accessible, with some available for free.

This follows closely on the heels of the latest generation of the Flash Thinking Experimental model (2.0), which introduces file uploads and an expanded 1 million token context window for Gemini Advanced users. This allows the model to analyse more complex information, such as legal documents or scientific articles, and perform tasks such as summarising key findings or comparing different perspectives. The model is also becoming more efficient and better at reasoning, which enables it to break down intricate user prompts into structured steps for more precise responses.

"Thinking about applying to law school? I'll help you weigh the pros and cons, compare programs, and write a compelling personal statement to submit with your application by Monday morning," Gemini responds to a hypothetical user query in a blog post by Google.

"Or, if you prefer food-related tasks, I can create a cookie recipe in the style of a famous chef, making sure it's gluten-free and highlighting any substitutions that might be needed."

Greater Efficiency and Deeper Research

The latest Experimental model is also powering a major upgrade to Deep Research, which was first introduced in December as a tool that could sift through vast amounts of information online and complete tasks in minutes.

Now, it's capable of performing multi-stage research tasks and compiling detailed multi-page reports. At the same time, users will get a real-time view of how Gemini is thinking through research queries, offering transparency into how it's finding and organising information.

"If you're planning a trip to Italy in December and want to find an apartment rental in Rome for a week, I'll search the web, compare prices on different travel booking sites, and summarise the best rental options for you to choose from," Gemini continues. "I'll even create a simple itinerary with restaurant suggestions and nearby attractions."

This upgraded version of Deep Research is rolling out globally in 45+ languages and is available in limited capacity for standard users and more broadly for Gemini Advanced subscribers.

Personalised AI Recommendations and Adaptability

Moreover, Gemini is evolving to become a more personalised assistant that can draw insights from a user's Google Search history (with their permission). This means suggesting restaurants based on past searches, offering travel advice tailored to a user's interests, and providing recommendations for relevant recipes or articles.

"If you've been researching French Impressionist painters and like recipes from the Minimalist Baker, I'll suggest an article about a new exhibit of Seurat's paintings at the local museum and recommend a fig and ricotta tart recipe," notes Google.

Users will be able to enable or disable Personalization (Experimental) through the settings, maintaining control over their data and the degree of personalisation they wish to engage with.

Expanding Connectivity with Google Apps

Google is also expanding Gemini's ability to interact with other Google apps, now encompassing Calendar, Notes, Tasks, and Photos. This enables users to issue complex, multi-step commands that span multiple services.

For instance, a single request like "Find an easy cookie recipe, add ingredients to my shopping list, and locate an open grocery store nearby" will be executed seamlessly, scheduling the baking session in Calendar, populating a Task to pick up groceries, and suggesting recipes and recipes from nearby stores.

Integration with Google Photos is coming soon, allowing users to retrieve details like trip highlights or document expiry dates through AI-powered image analysis.

"Show me photos of our family trip to Venice last December" might yield a highlight reel of the best moments from the trip, while "What's the expiry date of the passport in my photos?" would scan through images to locate the passport and extract the expiry date.

Custom AI Assistants: Gems

One of the most anticipated additions is the rollout of Gems, a feature that lets users create their own custom AI assistants, whether it's a language tutor, meal planner, or workout coach.

Previously available to select users, these personalised AI tools are being made more widely accessible and can be created by providing instructions and optionally uploading files for reference.

For instance, to create a language tutor Gem, a user could provide articles on specific grammar topics, examples of idioms and slang, and set learning goals to guide the tutor's responses.

To make a meal planner Gem, they might upload family recipes and food preferences, and set dietary restrictions to guide the creation of weekly meal plans and grocery lists.

"I want you to be a language tutor. Help me learn Spanish. Pay attention to common idioms and slang used by native speakers. I'm setting a goal to speak at an intermediate level by December," a user might instruct the Gem.

"Or, if you prefer something food-related