Imagine describing a look to a friend — “barrel jeans that aren’t too baggy, maybe ankle length, and something with a subtle acid wash” — and getting a scrolling palette of pictures that matches that vibe, plus direct links to buy what you like. That’s the pitch behind Google’s latest AI Mode update: make image search behave less like a filter-driven, checkbox exercise and more like a back-and-forth conversation that mixes words and pictures.

Google has added a visual, conversational layer to AI Mode that lets you start with text, an uploaded photo, or both — and then refine the results naturally (for example, “show me more ankle length” or “make it darker and bolder”). The company says the feature will surface “rich visuals” that match the vibe you describe, and that each result is shoppable so you can follow through to a retailer. The rollout is limited to English-language users in the U.S. this week.

Under the hood, Google leans on its multimodal Gemini models — the update specifically references Gemini 2.5 — to “see” and reason about images alongside text. That lets the system pick up on subtle visual signals (secondary objects, fabrics, proportions) and then fan out multiple visual searches in parallel to recombine the best matches. In plain language: instead of only matching keywords and tags, the search uses an AI understanding of what each image contains and how those pieces relate to your spoken description.

Google also connects that visual understanding to its Shopping Graph and existing Lens/Image Search infrastructure, so results can include product details, prices, reviews and links to retailers — the kind of real-time commerce data that makes a visual result actionable, not just inspirational.

This feels like the natural next step for product discovery: people often know vibe and fit better than precise size or SKU. Conversational prompts let shoppers start broad and then narrow down, which mirrors how humans browse online and in stores. For sellers, it’s an opportunity — a better match between intent and inventory should increase click-throughs to product pages. For shoppers, it’s less fiddling with dropdowns and more free-form browsing. Google is explicit that the feature is meant to be “shoppable,” not just decorative.

Visual exploration isn’t only commerce. Interior designers, hobbyists, and creators can use the same conversational flow to iterate on a mood board: upload a photo of a room or garment and ask for variations, color palettes, or “more like this but lighter.” The blend of text and image gives a faster path from vague idea to visual shortlist.

Google’s move is powerful but not risk-free.

• Commercial bias: Because results can link directly to retailers and surface product details, there’s a commercial incentive to favor listings that pay, rank better, or are better optimized. That could skew what “matches” look like in practice.
• Mistakes from visual reasoning: AI can misinterpret images — confusing textures, misreading scale, or overlooking less common design cues. That’s especially important when people rely on subtle distinctions (e.g., fabric type, fit).
• Privacy and image use: Uploading photos or snapping items in public raises questions about what Google stores, how long images are kept, and whether data will be used to improve models. Google’s announcement reiterates safety and product policies broadly, but users should be aware of the usual trade-offs when handing over images.

Visual search has been evolving for a while — Pinterest Lens, Microsoft’s Bing visual search and other apps have similar instincts — but Google’s advantage is scale: its Shopping Graph, product index, and Lens/Image Search history give it a head start in mapping images to commerce inventory and contextual web data. Whether that translates into more accurate, useful results will come down to execution and how Google balances relevance with commercial interests.

If you’re in the U.S. and use English, check AI Mode in Search — the new features should appear progressively over the week. Try starting with a loose, conversational prompt (the way you’d describe something in real life) and then use follow-ups to steer results. Flip between text-only prompts and the mixed image+text approach to see which gives the best matches for your needs.

This update is part of a broader trend: search engines trying to become more conversational and more visual at once. Google’s experiment is sensible — people think in pictures and words — but the payoff depends on whether the system really understands nuance and whether the results stay helpful rather than promotional. For shoppers and visual browsers, it promises speed and serendipity; for everyone else, it’s another step toward a search experience that responds like a human conversation — for better and, occasionally, for messier.