Pinterest AI for Ads and Discovery: Business Assistant Benchmarks, MCP Integrations and Performance+ Gains

Business Assistant: AI Embedded Directly in Advertiser Workflows

Pinterest’s Business Assistant is currently in closed beta in the United States, accessible within Ads Manager and on mobile. The core premise is straightforward — rather than requiring advertisers to interpret raw data independently, Business Assistant acts as an AI collaborator that surfaces actionable insights in context.

What distinguishes it from generic AI chat interfaces is its visual-first output. Instead of returning text summaries, it renders trend data as graphs and surfaces top-performing Pins directly. If searches for “clean beauty routine” spike 42% in a given week, Business Assistant visualizes that growth curve and presents the Pins driving it — giving advertisers both the signal and the creative reference in a single view.

The mobile extension adds a proactive dimension. Advertisers receive notifications about emerging trends, campaign performance shifts, and optimization opportunities without needing to actively query the system. This moves the tool from reactive assistant to ambient advisor.

Pinterest MCP: Standardized AI Infrastructure for Partner Ecosystems

The Model Context Protocol integration is arguably the most technically significant announcement in this release. Pinterest MCP provides a standardized, secure interface that connects Pinterest’s campaign data, analytics, and keyword insights to external copilots and agentic tools already in use by agency and brand teams.

The protocol is built on the open Model Context Protocol standard, which means it is designed for interoperability rather than lock-in. Alpha partners currently shaping its development include PMG, Pacvue, Dentsu, Havas, Innovid by Mediaocean, and Omnicom’s Jump450 — a roster that spans independent performance agencies, global holding companies, and ad tech platforms.

Chris Ivey, President of Jump450, framed the practical value clearly: the integration allows teams to analyze Pinterest performance, uncover insights, and act on opportunities without switching tools. The underlying advantage Pinterest brings to this infrastructure is its intent signal quality — users arrive on Pinterest actively planning future actions, which generates a distinctly high-value data layer for AI-driven campaign decisions.

Performance+ Creative: Asset-Level Optimization with Measurable Lift

Pinterest Performance+ creative has received a new AI selection model that shifts optimization from the ad level to the individual asset level. In practical terms, this means the system evaluates a broader pool of creative variants and selects the one most likely to perform for each specific ad impression — rather than applying a single winning variant uniformly.

The benchmark result from internal testing is a 7.5% increase in click volume compared to the previous single-variant model. While a single metric does not tell the complete story, a 7.5% click volume lift at scale represents a meaningful efficiency gain for advertisers running high-volume campaigns.

Alongside the new selection model, Pinterest is introducing enhanced ad review tools and more granular creative reporting breakouts. These additions address a persistent gap in creative optimization workflows — the ability to understand not just which creative performed, but how and where it appeared.

Ask Pinterest: An Experimental Surface for Conversational Commerce

Ask Pinterest is a separate, limited-access application designed to test conversational and agentic shopping experiences outside the core Pinterest product. It draws on Pinterest’s Taste Graph — the proprietary model mapping user preferences, intent signals, and aesthetic affinities — to power multi-step, context-retaining discovery sessions.

The use cases Pinterest highlights are deliberately complex: planning a dinner party within a budget, finding a genuinely personal gift, furnishing a room incrementally over time. These are decisions that do not resolve in a single search query, and they represent a category of consumer intent that current search interfaces handle poorly.

Ask Pinterest is explicitly framed as a learning environment. Insights from user behavior within the app will feed back into the main Pinterest product — a sensible approach that limits risk while accelerating experimentation with AI-native interaction patterns.

Reading the Strategic Signal

Taken together, these four announcements describe a coherent architectural direction. Pinterest is simultaneously building AI into its own advertiser-facing tools, opening its data infrastructure to external AI systems via MCP, improving the performance mechanics of its ad product, and creating a separate surface to test the next generation of AI-driven commerce.

The intent signal advantage Pinterest repeatedly references is not incidental marketing language. It reflects a genuine structural difference: a platform where users arrive with forward-looking purchase intent, curating ideas for things they plan to do or buy. That signal quality is what makes Pinterest’s data valuable as a grounding layer for AI workflows — and it is the asset Pinterest is now actively packaging for the agentic era.

For AI tool observers, the Pinterest MCP rollout is the development worth tracking most closely. As agentic workflows become standard practice in marketing teams, the platforms that publish reliable, standardized data interfaces will gain disproportionate presence in AI-assisted decision-making. Pinterest has moved early on this front, and the partner list suggests serious enterprise adoption is already underway.

The shift from keyword search to contextual recommendation is not a future scenario — it is the current competitive terrain. Pinterest’s Cannes 2026 announcements make clear that the platform intends to be infrastructure, not just inventory, in that new environment.