Top Web AI updates from Google I/O 2025

Key Concepts:

• Chrome AI APIs: Prompt API, Summarize API, Language Detector API, Translator API, Writer API, Rewriter API, Proofreader API.
• Gemini Nano: Google's on-device AI model.
• Multimodal AI: AI that can process multiple types of data (text, image, audio).
• Hybrid SDK: Extends API reach to more browsers and devices.
• Origin Trial: A program to test new web platform features.
• AI Agent-Compatible Websites: Websites designed to interact with AI agents.
• Client-Side Processing: Processing data directly in the user's browser.

I. API Updates and Availability

• Four APIs Shipped: The Prompt API (for Chrome extensions), Summarize API, Language Detector API, and Translator API are now available. The Prompt API for web remains behind a flag.
• Two APIs in Origin Trial: The Writer and Rewriter APIs are being moved into origin trial for further testing and feedback.
• Three New APIs (Behind a Flag): The Prompt API with image input, the Prompt API with audio input, and the Proofreader API are available for testing behind a flag.
• Hybrid SDK Developer Preview: A hybrid SDK developer preview has been announced to extend the reach of these APIs to browsers and devices that wouldn't otherwise be covered.

II. Real-World AI Feature Examples

• The presentation focuses on practical applications of the new Chrome AI APIs rather than theoretical discussions.

III. Vision Nanny: A Case Study

• Description: Vision Nanny, part of the Google for Startups program in India, is a web platform designed for children with cerebral visual impairment.
• Significance: It demonstrates the potential of these APIs to address real-world problems and improve accessibility.
• Client-Side Processing: The use cases run fully client-side, enabling access anywhere, anytime.

IV. Chrome Extensions and Gemini Nano

• Gemini Nano Integration: Chrome's new built-in AI APIs allow developers to use Gemini Nano on the client-side.
• Impact on Developers: This integration unlocks new ways to customize and optimize browsing experiences for users.
• Exciting Time for Chrome Extension Developers: The combination of Gemini on-device and in the cloud creates a significant opportunity for innovation.

V. Adobe Acrobat Extension: Document Analysis

• Problem: Many businesses still rely on printed documents, making information retrieval slow and manual.
• Solution: Adobe experimented with the Prompt API and multimodal support in their Acrobat extension to analyze historic stock certificates with faded text.
• Process:
  1. The extension analyzes scanned visuals.
  2. Users can generate text summaries from the scanned documents directly within Chrome.
  3. Users can chat with the Acrobat AI assistant to surface, organize, and validate key facts.
• Benefits: Reduces manual work and enables faster, more accurate capture of critical information from sparse or degraded documents.

VI. AI Agent-Compatible Websites

• Prediction: Websites of the future may need to be AI agent-compatible.
• Demonstration: A web AI prototype using Google's two-billion parameter Gemma 2 model (running locally in the browser) is demonstrated.
• Concept: Users can interact with websites naturally via text or voice.
• Browser Advantage: The browser is an ideal platform because users are already signed in to various services, enabling access to tools from those services in one place.
• Call to Action: Start exploring how to apply web AI agents to your industry.

VII. Getting Started with the New Tools

• Availability: Many of the APIs are broadly available.
• Early Preview Program: Sign up for the early preview program to start experimenting with the new multimodal AI and hybrid solutions.

VIII. Conclusion

The presentation highlights the significant advancements in Chrome AI APIs, particularly the integration of Gemini Nano and multimodal capabilities. These advancements empower developers to create innovative solutions, improve accessibility, and streamline workflows. The emphasis on real-world examples and the call to action encourage developers to explore the potential of these tools and contribute to the future of the web.