Pinterest rolls out AI ad tools and an experimental shopping app

Planning a budget dinner party or furnishing a living room piece by piece usually turns into a chaotic mess of tabs and saved pins. Pinterest thinks the fix isn't another generic search bar, but an AI that already knows your aesthetic.

Ahead of Cannes Lions 2026, the company rolled out a batch of AI updates headlined by Ask Pinterest, a limited-access experimental web app that pairs conversational shopping with its proprietary Taste Graph.

Chatbots with taste, and visual ad helpers

Ask Pinterest lives outside the flagship app for now, accessible via mobile and desktop web for selected testers. Once you sign in, it pulls from your saved Pins and Boards to understand what you actually like, keeping context across multiple sessions instead of treating every query like a blank slate.

Unlike standard text-heavy chatbots, it leans heavily into Pinterest’s visual-first DNA to guide multi-step shopping journeys.

On the business side, Pinterest rolled out Business Assistant in a closed US beta across desktop Ads Manager and mobile. It skips the usual AI walls of text, offering trend charts and top-performing Pins instead. If searches for "clean beauty routine" spike 42% in a week, the assistant flags the surge, graphs the trajectory, and shows which creative assets are winning.

Meanwhile, a new dynamic asset-selection model inside Pinterest Performance+ creative evaluates individual asset pieces on the fly. In tests, picking the right creative variant for each individual impression lifted click volume by 7.5% compared to the older singular model.

Connecting to third-party ad agents

Perhaps the most practical addition for marketers is Pinterest MCP (Model Context Protocol). It is an AI-native integration that pipes Pinterest’s campaign analytics, keyword insights, and trend signals straight into the custom copilots ad agencies already use.

Instead of forcing media buyers to log into yet another dashboard, the protocol feeds into tools from alpha partners including PMG, Pacvue, Dentsu, Havas, Innovid by Mediaocean, and Omnicom's Jump450.

Jump450 President Chris Ivey noted that the integration lets their teams spot opportunities directly within existing workflows, tapping into Pinterest's high-intent planning signals without switching tools.

Why it matters

While Google and OpenAI rush to turn chatbots into universal shopping engines, Pinterest is playing a much sharper hand.

Spinning out Ask Pinterest as a sandbox is smart. It lets the team test complex, conversational commerce without alienating the millions of users who just want to scroll an aesthetic feed in peace. If you are decorating an apartment or planning an event over three months, a conversational tool with real visual context is genuinely useful; for quick commodity buying, it probably is not.

Strategically, Pinterest is refusing to become just another dataset licensed out to third-party LLMs. By hoarding its Taste Graph and only exposing insights via standardized interfaces like MCP, it keeps agencies locked into its ecosystem.

We think this shows where commercial AI is actually heading. Generic LLMs still struggle with personal taste, and by putting its unique intent signals inside both consumer sandboxes and enterprise agent protocols, Pinterest is making sure it stays relevant no matter where the search-and-click model goes next.