Google AI Updates: April 22, 2026

1. Google Ships Deep Research and Deep Research Max Agents on Gemini 3.1 Pro

Google. Google opened a public preview of two research agents on the Gemini API: Deep Research, optimized for low-latency chat-style use, and Deep Research Max, which trades latency for thoroughness and is aimed at overnight asynchronous runs. Both sit on Gemini 3.1 Pro and support Model Context Protocol connectors to proprietary data sources, multimodal inputs (PDFs, CSVs, images, audio, video), native chart and “Nano Banana” infographic generation, and optional web-access restriction. Availability is paid Gemini API tiers now, with Google Cloud rollout for startups and enterprises planned. Source

2. Ads Advisor Gets Three Agentic Safety Features for Google Ads

Google. Ads Advisor, Google Ads’ in-product AI agent, picked up proactive troubleshooting that flags complex policy violations before appeals, a 24/7 security-monitoring loop with a dedicated protection dashboard, and instant certifications that decide whether a business needs a certificate and either auto-approves or streamlines the application. The launch positions the three features as moving Ads Advisor past creative-idea generation into operational account hygiene. Source

3. Gemini in Chrome Expands to Seven APAC Markets

Google. Gemini in Chrome rolled out to Australia, Indonesia, Japan, the Philippines, Singapore, South Korea, and Vietnam. Desktop and iOS are supported everywhere except Japan, which gets desktop only. The package includes sidebar Q&A across tabs, Personal Intelligence tying Gmail, Photos, Maps, and Calendar together, and Nano Banana 2 image transformation — though agentic task automation remains US-only and limited to AI Pro and Ultra subscribers. Source

4. YouTube Extends AI-Likeness Detection to Celebrities Through Talent Agencies

Google. YouTube opened its likeness-detection system — a face-matching scan modeled on Content ID — to entertainment industry talent. Agencies including CAA, UTA, WME, and Untitled Management are the launch partners, and importantly the feature doesn’t require enrollees to have their own YouTube channels. Flagged videos yield three response options: removal request for privacy violations, copyright claim, or no action, with explicit carve-outs for parody and satire. Audio detection is next on the roadmap. YouTube also restated support for the federal NO FAKES Act. Source