Pinterest this week announced four distinct AI infrastructure products ahead of Cannes Lions, including a Model Context Protocol server that connects the platform to external advertising tools, an in-Ads-Manager AI collaborator in closed beta, an updated creative selection model that lifted click volume by 7.5% in testing, and a standalone experimental app called Ask Pinterest designed for conversational shopping.

The announcement, published on June 17, 2026, positions Pinterest's infrastructure layer - rather than its consumer product - as the headline story. That is a notable shift. For most of its history, Pinterest has led advertising announcements with new ad formats or targeting capabilities. What the company is presenting ahead of Cannes Lions 2026 is something closer to plumbing: standardized interfaces for AI agents, mobile-first workflow tooling for advertisers, and a separate experimental app that serves as a testing ground for the next generation of shopping experiences. The mix of products is technically varied, but each component addresses a different layer of the same problem - how a discovery platform integrates into a marketing ecosystem increasingly organized around AI agents and conversational interfaces.

Pinterest enters Cannes Lions 2026 with 631 million monthly active users and $1.008 billion in Q1 2026 revenue, an 18% year-on-year increase. The financial backdrop matters for reading these product announcements: the company is not rebuilding a struggling platform, but extending infrastructure built on a foundation that is already monetizing. Performance+ campaigns, which Pinterest launched in October 2024, already drove 30% of lower-funnel revenue by Q1 2026.

Business Assistant: an AI collaborator inside Ads Manager

The first product is Business Assistant, an AI collaborator embedded in Pinterest's Ads Manager and, for the first time, in a mobile interface. Pinterest describes it as currently in a closed beta in the United States. The system builds on an earlier product called Pinterest Assistant, which launched in beta in Q4 2025 and remained in controlled rollout through Q1 2026.

What distinguishes Business Assistant from a generic AI chatbot is its visual output layer. Rather than returning text responses, the tool displays trend data as graphs and shows actual Pins as examples. According to Pinterest, if searches for "clean beauty routine" rise 42% in a given week and an advertiser wants to capitalize on that movement, Business Assistant represents the interest curve graphically alongside the real Pins that are leading the trend. That approach - grounding trend recommendations in visual examples rather than descriptive text - reflects Pinterest's position as a platform where the image itself is the unit of commercial communication.

The mobile dimension adds a proactive element. Advertisers receive notifications about emerging trends, performance status, and optimization opportunities without actively querying the tool. That shifts the product from reactive assistant to proactive monitor.

The advertiser workflow implications are practical. Campaign managers at agencies or in-house teams can query performance, receive trend alerts, and pull actionable recommendations without leaving the Ads Manager environment or switching between tools. Pinterest's Media Planner, launched January 23, 2026, eliminated spreadsheet-based upper-funnel planning by embedding audience discovery and budget estimation directly inside Ads Manager. Business Assistant extends that consolidation principle into the performance analysis and trend monitoring layer.

What "closed beta in the US" means for the broader market

Closed betas on advertising platforms follow predictable patterns. A small group of US advertisers - typically large spenders or strategic partners - receive access first. Feedback from that cohort shapes the product before broader rollout. There is no stated timeline in the announcement for when Business Assistant moves beyond the closed beta phase, and non-US advertisers have no confirmed access date.

Pinterest MCP: standardized AI agent access to campaign data

The second product is Pinterest MCP, a Model Context Protocol server that allows external AI tools and copilots to connect directly to Pinterest's campaign data, analytics, and keyword insights. This is the announcement with the most significant structural implications for the marketing technology ecosystem.

Model Context Protocol is an open standard that Anthropic developed in November 2024 and later donated to the Linux Foundation. It functions as a standardized communication layer between AI applications and external data sources - a consistent interface that allows large language models to query APIs, databases, and tools without requiring custom integrations for each connection. PPC Land has tracked MCP adoption across the ad tech sector since the standard's launch, covering implementations from Google, Meta, Amazon, Yahoo DSP, FreeWheel, AdRoll, PubMatic, and SISTRIX, among others.

Pinterest's implementation connects three categories of data through the MCP interface: campaign performance data, analytics, and keyword insights. The protocol also exposes Pinterest's proprietary signals - including the Taste Graph, which is trained on billions of images, videos, and text - to partner tools. That last element is where Pinterest's MCP offering differs from other ad platform implementations. Most MCP servers in advertising expose campaign management data. Pinterest's offering additionally surfaces intent and taste signals that are specific to its platform and are not replicated elsewhere.

The alpha partner list named in the announcement includes PMG, Pacvue, Dentsu, Havas, Innovid by Mediaocean, and Omnicom's Jump450. These are not small players. PMG is already piloting FreeWheel's MCP server through its Alli marketing operating system. Pacvue is a major retail media and commerce management platform. The inclusion of Dentsu and Havas signals that Pinterest is targeting agency holding company integration, not just technology platforms.

According to Chris Ivey, President of Jump 450: "Pinterest MCP helps us integrate Pinterest directly into the workflows our teams are already building, making it easier to analyse performance, uncover insights and act on opportunities without switching tools. What makes Pinterest especially compelling is the strength of its intent signals. People come to Pinterest to plan what they want to do next, creating valuable context that helps marketers make smarter decisions."

The practical consequence of MCP integration is meaningful. Before MCP, accessing Pinterest campaign data from an external tool required either manual export or custom API development by engineering teams. With an MCP server, a partner copilot - whether that is Pacvue's interface, an agency's internal tool, or an AI assistant running inside a planner's workflow - can query Pinterest's data through a standardized interface with authentication handled by Pinterest's own systems. AdRoll and PubMatic demonstrated in April 2026 what that kind of inter-platform agent communication can accomplish: deal diagnostic resolution that previously took days was compressed to minutes once agents could query across platform boundaries without human intermediation.

Pinterest MCP is described as evolving alongside its alpha partners, with feedback from those partners shaping the protocol across reporting, analysis, campaign planning, and execution. There is no stated date for general availability.

Protocol context: where Pinterest MCP sits in a crowded field

The MCP adoption curve in advertising has moved quickly. Google released an open-source MCP server for its Ads API on October 7, 2025. Amazon launched a closed beta for its own MCP server on November 13, 2025. Meta opened its advertising infrastructure to Claude and ChatGPT through MCP in late April 2026 - the first time Meta had opened its ad system to third-party AI at that level of integration. Yahoo DSP launched MCP-enabled agentic capabilities on January 6, 2026. DoubleVerify embedded MCP into its Open Connectivity pillar this week, alongside the Pinterest announcement.

Pinterest's MCP arrives in a market where the protocol is no longer experimental infrastructure - it is becoming standard expected capability for any advertising platform that wants to participate in agent-mediated campaign workflows. PPC Land's coverage of the agentic advertising stack this week noted that the buying, measurement, and discovery layers of digital advertising are all moving toward agent-accessible interfaces simultaneously.

Pinterest Performance+ creative: dynamic asset-level optimization

The third product is an updated AI model for Pinterest Performance+ creative, the creative automation layer that sits inside the Performance+ suite. The existing product, which Pinterest has been developing since the October 2024 global launch of Performance+, already used Pinterest's Canvas AI image model to automatically edit product image backgrounds and transform catalog images into lifestyle imagery.

The new model changes how creative selection operates. Where the previous system used a single variant approach, the updated model evaluates a broader set of creative assets and identifies which variant is most likely to perform for each individual ad impression. According to Pinterest, testing of the new model produced a 7.5% increase in click volume compared to the prior single-variant model.

That shift - from ad-level optimization to asset-level optimization - has a meaningful structural implication. Instead of a campaign performing as a unit, the system evaluates individual creative assets against individual impressions and selects the combination most likely to convert. Advertisers with larger creative libraries gain a direct performance advantage, because the model has more variants to evaluate. Conversely, advertisers with limited creative assets have fewer options for the model to select from.

Pinterest Performance+ creative with the new AI model is available to advertisers globally. The announcement also references new ad review tools and enhanced creative reporting breakouts - the details of which are not elaborated in the announcement, but which address a recurring complaint among advertisers on automated creative platforms: limited visibility into how individual assets are performing.

The Performance+ context matters here. Pinterest launched the full Performance+ suite in October 2024, positioning it as a collection of AI and automation tools for campaign performance. By Q1 2026, Performance+ drove 30% of lower-funnel revenue. Advertisers using Performance+ were growing their lower-funnel spend at nearly twice the rate of non-adopters. That adoption curve gives the updated creative model a large and growing base of campaigns into which the new selection logic will deploy.

Ask Pinterest: an experimental conversational shopping app

The fourth product is the most structurally distinct. Ask Pinterest is a new, separate application - not a feature inside the existing Pinterest app - designed for conversational, visual-first shopping experiences. Pinterest describes it as rolling out in the United States with limited access.

The product draws on Pinterest's Taste Graph, the multimodal AI system trained on billions of images, videos, and text that powers Pinterest's recommendation engine, and which Pinterest recently committed $4 billion to AWS through 2031to train more cheaply on Amazon's Trainium chips. Ask Pinterest applies those signals - taste, intent, and stated preferences - to generate recommendations through a conversational interface rather than the standard browse-and-save flow.

The use cases that Pinterest identifies as targets for Ask Pinterest are instructive: planning a dinner party on a budget, finding a gift that feels personal, furnishing a room over time. These are multi-step, context-dependent decisions that unfold over multiple sessions. They do not fit the single-query search model. And they are precisely the category of shopping decisions that conversational AI handles better than keyword search: the user's intent evolves during the conversation, prior context is retained across sessions, and the output is a curated shortlist rather than a results page.

Ask Pinterest is explicitly described as an experiment. What Pinterest learns from its behavior - how users interact with conversational recommendations, where the experience breaks down, what retention looks like across sessions - will feed back into the main Pinterest app's development. That framing positions Ask Pinterest as a fast-iteration vehicle for features that, if validated, would eventually integrate into the 631-million-user platform.

That approach mirrors the product development logic that Pinterest CEO Bill Ready described on the Q1 2026 earnings call when discussing Pinterest Assistant: deliberate pacing, learning before scaling, avoiding the false starts that other platforms experienced when they moved too quickly from beta to wide deployment.

The broader agentic shopping context

Ask Pinterest does not exist in isolation. The Cannes Lions week that Pinterest is using for its announcement is the same week that the advertising industry is processing the implications of AI agents becoming the primary interface for consumer discovery. PPC Land's analysis of the week noted that WPP Media projects global advertising will reach $1.3 trillion in 2026, with AI infrastructure investment as the primary driver. Ask Pinterest is Pinterest's attempt to be part of that interface layer rather than content surfaced through it.

The distinction matters. A platform that is merely indexed by AI agents is a content supplier. A platform with its own conversational interface, backed by proprietary taste signals, can potentially capture user intent directly and route it toward advertiser-connected purchase paths - without the platform's recommendations being filtered through a third-party agent. That is the strategic position Ask Pinterest is designed to establish, even if it currently serves a limited number of US users.

Why the marketing community should pay attention

Pinterest today is announcing four products. None of them is a new ad format. Three are infrastructure components (Business Assistant, Pinterest MCP, the Performance+ creative model) and one is an experimental consumer interface (Ask Pinterest). That composition reflects where the platform is directing engineering investment: the backend systems that connect Pinterest to external marketing workflows, and the front-end experience that may define how conversational AI handles visual shopping.

For media buyers, the Pinterest MCP announcement is the most immediately actionable of the four. If partners like PMG, Pacvue, or agency-developed tools integrate Pinterest's MCP server, buyers at those organizations will gain the ability to query Pinterest campaign data and receive Pinterest-specific recommendations without leaving existing interfaces. That changes how Pinterest competes for share of attention in multi-platform planning workflows - not by demanding that planners learn a Pinterest-specific interface, but by embedding Pinterest signals into whatever tool those planners already use.

For creative teams, the updated Performance+ creative model changes the economics of creative production. A 7.5% click volume lift from dynamic asset selection rewards investment in creative diversity. Producing 10 creative variants instead of three is not just a brand preference - it becomes a measurable performance input.

For platform strategy watchers, Ask Pinterest is the signal to track. The no-experience Promote a Pin that launched in March 2026 extended the advertiser base downward. Ask Pinterest extends the consumer interface outward. Together they describe a platform expanding simultaneously in both directions: toward simpler advertiser entry and toward more sophisticated consumer discovery.

Timeline

Summary

Who: Pinterest, the visual discovery platform with 631 million monthly active users, made these announcements. Named individuals include Lee Brown, Chief Business Officer at Pinterest, and Chris Ivey, President of Jump 450, one of the alpha partners for Pinterest MCP. Alpha MCP partners named are PMG, Pacvue, Dentsu, Havas, Innovid by Mediaocean, and Omnicom's Jump450.

What: Pinterest announced four AI products on June 17, 2026: Business Assistant, an AI collaborator in Ads Manager and mobile currently in closed beta in the US; Pinterest MCP, a Model Context Protocol server providing alpha partners with secure access to campaign data, analytics, keyword insights, and Pinterest's Taste Graph signals; an updated Performance+ creative AI model that increased click volume by 7.5% in testing through dynamic asset-level selection, available globally; and Ask Pinterest, a separate experimental app for conversational visual shopping rolling out in the US with limited access.

When: The announcement was published on June 17, 2026, one week before Cannes Lions 2026, which runs June 22-26.

Where: Business Assistant is in closed beta in the United States. Pinterest MCP is in alpha with named agency and technology partners. The Performance+ creative model is available to advertisers globally. Ask Pinterest is rolling out in the United States with limited access.

Why: Pinterest is building infrastructure to connect its platform to the external AI agent workflows that advertising teams are adopting, while simultaneously building a consumer-facing conversational interface that applies Pinterest's proprietary taste and intent signals beyond the core app. The timing ahead of Cannes Lions is deliberate: the festival has become the primary venue where the advertising industry evaluates platform infrastructure commitments for the following twelve months.