Google Search's AI Mode goes fully visual

We've all been there: you have a crystal-clear image in your head of an apartment aesthetic or a fall jacket, but you sit staring at a blank search box because words just completely fail you.

Google just rolled out a major upgrade to AI Mode in Google Search that aims to fix exactly that, turning vague visual vibes into interactive, conversational searches.

Ditching filters for visual conversation

Starting this week for English users in the U.S., you can ask conversational questions and get a rich grid of visual inspiration instead of plain links or walls of text.

Say you're redesigning your bedroom and want maximalist decor. You tell AI Mode the general vibe, get a set of matching visuals, and then naturally refine it by asking for "more options with dark tones and bold prints." Every image links out to the source, and on mobile, you can even tap inside a specific image to ask follow-up questions about an item in the corner.

The same conversational flow carries right into shopping. Instead of toggling ten different drop-down filters for cut, wash, and rise, you can literally type "barrel jeans that aren’t too baggy" and follow up with "show me ankle length." Google surfaces buyable items backed by its Shopping Graph, which houses over 50 billion product listings, refreshing more than 2 billion of them every single hour to keep prices and stock accurate.

Under the hood, Google is pairing Google Lens visual search with Gemini 2.5's multimodal chops. The secret sauce is a new technique called "visual search fan-out." Rather than just identifying the main subject of an image, the model analyzes subtle background details and secondary objects, firing off multiple queries in the background simultaneously to match your natural language intent.

Why it matters

This looks like a genuinely smart pivot for Google Search. While most AI search startups are obsessing over summarizing text and citing academic papers, Google is leaning heavily into visual discovery and commerce—areas where text-only prompts fall completely flat.

For everyday users and shoppers, this solves a real headache. Trying to filter through traditional e-commerce categories using rigid keywords is tedious; telling an AI what you mean like you're chatting with a stylish friend is a massive quality-of-life upgrade that is immediately worth trying for home decor or fashion.

For competitors, this is a flex of Google's deepest moat. Startups can easily license multimodal models, but they don't have a 50-billion-item Shopping Graph refreshing 2 billion products an hour. Google is reminding everyone that real-time infrastructure and commercial index depth still matter just as much as model intelligence.