The 2025 Google I/O keynote could easily be dubbed “The Google AI Showcase.” Almost every announcement revolved around advancements powered by artificial intelligence, featuring innovations that are either available now or set to debut in the near future. Below are the standout features across Google’s product ecosystem that are likely to be of interest.

Understanding Gemini

Delving into Gemini can be somewhat complex, as it encompasses various models, including Gemini Flash, Gemini Pro, and Gemini Pro Deep Research. Different iterations of these models, like version 2.5, are also included, alongside several applications where these capabilities are utilized. You can find Gemini integrated into the dedicated Gemini app, as well as voice assistants embedded in devices like Pixel phones and smartwatches. Additionally, it enhances platforms such as Google Docs, Gmail, and Search.

Introducing Agent Mode

One of the major highlights is the arrival of Agent Mode in the Gemini app, which allows it to handle tasks while you focus on other activities. During the event, Google showcased an example where users could instruct Gemini to search for apartment listings in their preferred city. The app filters options based on predefined criteria and can even arrange tours for interested users.

This feature is designed to perform repetitive searches seamlessly. For instance, if users want Gemini to find new apartment listings each week, it can carry on with the process, utilizing information from previous searches.

Similarly, this capability will be integrated into Google Search for specific inquiries. For example, asking for event tickets would prompt Google to sift through ticketing websites, compare them with user preferences, and deliver tailored results.

Gmail’s Smart Replies Enhanced

Gmail has long included smart replies, but they often lack personalization, giving away that you aren’t fully engaged in the conversation. Soon, Gmail will enhance its response suggestions by analyzing your past emails and documents stored in Drive. This means it can draft replies reflecting your typical style and advice.

Consider a scenario where a friend inquires about your recent trip planning. Gmail will create a tailored response based on your previous interactions, ensuring the tone and details are characteristic of your writing style.

Summarizing AI Thought Processes

A fascinating new feature involves recapping the AI’s thought process. Typically, AI models work by decomposing queries into smaller tasks and processing each step. Gemini traditionally displays these steps, but for those seeking more clarity, it will now provide a concise summary of its reasoning process. This streamlining aims to clarify the logic behind AI-generated answers.

New Audio Features

Another intriguing addition is the introduction of native audio output via the Gemini API. Developers can create applications that utilize natural-sounding voices. A notable demonstration showed off the ability for voices to switch languages seamlessly. However, one unexpected twist is that the model can also whisper. The potential applications for whispering AI voices remain unclear, but it certainly adds an element of intrigue.

Jules: The Coding Assistant

Last year, Google introduced Jules, a coding assistant akin to GitHub’s Copilot. The public beta is now accessible, and Google asserts that Jules can autonomously address bugs while users engage in other projects. Aside from debugging, it can also update dependency versions and provide audio summaries of the modifications it implements.

Virtual Try-Ons for Online Shopping

A new feature in Google Search allows users to virtually try on clothing while shopping online. By uploading a full-length picture, Google will showcase how the apparel would look on your body type. Additionally, upcoming shopping tools will enable price tracking and purchasing through Google Pay, utilizing saved payment and shipping information. While this feature is not yet available, further clarity on its functionality and safeguards for unwanted transactions will be necessary before widespread usage.

Innovations in Video and Audio Generation

Google also unveiled new models, Veo and Imagen, aimed at generating audio and video content. The demonstration of Veo 3 effectively produced video, although its quality will depend on user perception. The company appears to be banking on the appeal of the content created by Veo 3 and its associated images from Imagen 4, accompanying the announcement of a video editing suite named Flow. This tool reportedly enables editors to extend and regenerate clips for optimal results, complete with sound effects that align with the visuals. Veo 3 is set to be available in the Gemini app for Ultra subscribers.