Google’s AI Mode gets eyes

Visual search used to mean snapping a picture of a sneaker to find a store link. Google now wants its search engine to actually understand the full context of what you are looking at.

Google rolled out multimodal visual capabilities to its AI Mode search experience, while expanding access to millions more Labs users in the US across the Google app on Android and iOS.

Combining Lens with fan-out queries

The upgrade merges a customized Gemini model with Google's Lens visual recognition technology. Users can snap or upload an image, and AI Mode returns a comprehensive response complete with web links about what is inside the frame.

According to Robby Stein, VP of product for Google Search, the feature goes beyond identifying isolated items. With Gemini, AI Mode can interpret an entire scene, understanding spatial relationships between objects, materials, shapes, and layouts.

To make answers more helpful, Google is using what it calls a fan-out technique. When processing an image, the system automatically fires off multiple parallel search queries behind the scenes about the various elements it detects.

For instance, take a picture of a shelf of books, and the tool can identify each title, suggest similar books with strong review scores, and answer follow-up prompts to narrow down your reading list. Google noted on X that users are increasingly leaning on AI Mode for open-ended, complex tasks like in-depth comparisons and how-to guides.

Moving out from behind the paywall

AI Mode is Google’s direct counterpunch to conversational search tools like Perplexity and ChatGPT Search. Instead of a list of blue links, it delivers AI-curated overviews grounded in Google's search index.

When AI Mode debuted last month, Google locked it behind its Google One AI Premium subscription tier in Labs. Opening the feature up to millions of non-paying Labs users in the US shows the company is eager to gather broader real-world data and defend its search turf.

Why it matters

We think combining Lens with Gemini gives Google a clear edge over pure text-first AI search engines. Perplexity and OpenAI have strong reasoning, but Google has a decade of computer vision infrastructure and the world's most up-to-date web index.

For regular users, this turns visual search into a genuinely useful research assistant. Instead of trying to describe a strange mechanical part or a room layout in words, you can simply upload an image and ask for step-by-step guidance.

The fan-out approach looks smart because it prevents the AI from guessing based on visual impressions alone by grounding each sub-component in web results. However, running multimodal reasoning paired with multiple background queries is computationally expensive, and keeping access limited to US Labs users suggests Google is still cautious about scaling the infrastructure globally.