Pinterest

Unified and retrained Shopping ROAS models to optimize ad stack performance.

Updated proprietary search ranking model, extending user context windows by 30-fold.

Acquisition of TV Scientific closed in Q1 to integrate audience data into CTV campaigns.

Global extension of PinRack generative retrieval system to serve content site-wide.

CEO Bill Ready articulated the strategy that fit-for-purpose visual search wins over general-purpose LLMs.

“As AI adoption accelerates, general purpose search is increasingly up for grabs, as the largest players pour capital into general purpose LLMs. But fit-for-purpose search still wins in key verticals like travel and consumer products.”— Bill Ready

Acquired TV Scientific to support monetization beyond owned properties, leveraging high-intent audience for connected TV performance advertising.

Management disclosed that roughly 50% of new code is AI-generated.

Anticipates approximately 100 basis points cost-of-revenue headwind in 2026 aligned with GPU spend for foundational AI models and Pinterest Assistant.

Pinterest Assistant launched in beta in Q4 2025.

Pinterest Performance Plus New Customer Acquisition entered beta, allowing advertisers to assign customized values to audiences.

Launched Pinterest Assistant in beta, a voice-activated, visual-first conversational AI tool.

Performance+ creative tools, including automated cropping, brightness adjustment, and logo overlays, expected to launch in 1H 2025.

ROAS bidding, which optimizes bids to maximize return on ad spend, scheduled for release to all eligible advertisers by end of Q1 2025.

Performance Plus ROAS bidding expected to exit beta and reach general availability in Q1 2025.

CEO Bill Ready emphasizes AI as a core competency for the visual search platform.

“As a visual search platform, AI is a core competency here at Pinterest.”— Bill Ready

Developed Navigator 1, a model framework leveraging visual embeddings that reduces latency and delivers a 90% cost reduction compared to third-party proprietary models.

Launched PinFM, a proprietary foundation ranking model that distills lifetime user actions into feed recommendations, driving a 240 basis point increase in saves.

Introduced OmniSage, a proprietary core AI model trained on the taste graph, driving a 450 basis point lift in sitewide saves.

Transitioned from CPU to GPU serving to leverage larger models for organic and ad serving.

15% of the company's current code base is generated through AI coding assistance tools.

Increased the context window on user actions fed into AI-based models by over 30-fold, resulting in 250 basis points of lift in saves and 150 basis points of lift in outbound clicks.

Pinterest's Taste Graph, the foundation of its AI recommendation engines, grew by 75% over the past two years to contain billions of connections.

Performance Plus adoption resulted in 50% fewer inputs for campaign setup and a 20%+ cost-per-action improvement for shopping campaigns.

Performance Plus Creative beta testing demonstrated a 14% average increase in conversion rate and a 9% lower cost per action.

Average ad relevance for top ads on the search surface has more than doubled over the past two years due to AI-powered ranking and delivery systems.

Fine-tuning off-the-shelf LLMs with first-party signal resulted in a 300 basis point improvement in actionable engagement.

CEO Bill Ready highlights AI as a core competency powering all content recommendations and ad serving.

“At Pinterest, AI is a core competency and has played a central role in how we have transformed our business over the last two-plus years for both our users and advertisers.”— Bill Ready

Performance+ ad product suite, which utilizes AI to automate campaign setup and improve performance, reached general availability.

General availability of Performance Plus Creative, featuring generative AI backgrounds for shopping ads.

General availability launch of Performance Plus, an AI-driven ad suite for budgeting, bidding, and targeting.

CEO Bill Ready outlines the vision for full AI-driven ad campaign automation.

“Our goal is for advertisers to be able to provide us with a budget, a goal, and their seed creative and we'll do the rest.”— Bill Ready

Performance Plus Alpha test participants saw a greater than 10% improvement in cost per acquisition for lower funnel conversion campaigns.

Introduced Performance Plus Creative, using generative AI to optimize and create ad assets.

Launched Performance Plus in beta, an AI/automation suite for bidding, budgeting, and targeting.

Extended AI-powered collages (computer vision) to advertisers for ad creation.

Launched AI-powered auto-organization for boards to automatically group Pins.

Upgraded core search ranking algorithm to incorporate new signals, improving relevance and personalization.

Company is increasing R&D headcount specifically for AI talent.

Number of users searching with skin tone ranges doubled in the past year, attributed to inclusive AI.

CEO Bill Ready on Pinterest's strategy of using unique platform signals to train fit-for-purpose models rather than relying solely on general-purpose models.

“I do think this first wave of AI or some of the first conversations in AI were about value accruing to the model creators and those who have like the largest general-purpose models. I think it's also the case that there's a distinct set of value creation around both fit-for-purpose models and taking unique signal applying it to those larger models, and you see us doing both.”— Bill Ready

Collages feature, powered by AI and computer vision, enables users to curate and piece together content.

Introducing ROAS bidding, an AI-driven tool to automatically optimize campaigns in real-time.

Launching dynamic creative optimization solution set which allows advertisers to use generative AI to optimize ad creative.

Expanded generative AI-based guided search for content discovery in the Home Decor vertical.