Pinterest Has 640 Million Users and a Working AI Shopping Tool. Growth Is Still the Problem.

Pinterest reported Q2 2026 results on August 4, 2026. In the call, they said that the AI shopping pivot is working with revenue up 18%, Performance+ carrying 30% of lower-funnel ad revenue, and Pinterest Assistant live for most US users. 

The core market growth is stalling with flat user additions in the US and Canada, two million users lost in Europe, and three consecutive quarters of mostly stagnant usage in the regions that generate the platform's highest revenue per user, according to the release.

The company is converting its existing 640 million user base into a more commercially valuable audience, ad revenue per user is rising, large retailers are spending more, and small businesses are adopting automated shopping campaigns. But it is not adding new users in the markets where that conversion is most financially meaningful.

"AI is at the heart of our momentum and is a clear accelerant for our business," said CEO Bill Ready. "It is trained on our unique human curation of style and taste, making Pinterest more personalized and actionable for users, while improving performance for advertisers and creating more opportunities to monetize over the long term."

Pinterest's AI strategy rests on a specific data asset. The company's Taste Graph, built from 80 billion monthly searches and 16 billion boards, is used to train AI models on individual user taste and intent rather than relying on general-purpose models, according to the press release. 

More than 96% of Pinterest's text-based searches are unbranded, which is the gap the company is building AI to close. Pinterest Assistant, launched to most US users in late July, is the consumer-facing expression of that strategy. 

The tool allows users to refine ideas, compare products, get personalized recommendations, and answer shopping questions within a conversational interface, targeting the research phase of the purchase journey rather than the browsing phase the platform has historically owned.

The cost architecture is as notable as the capability. Pinterest post-trains open-source models on its own data rather than using closed proprietary systems from external vendors, according to the press release. 

Ready disclosed the result directly: cost per transaction running at less than 8% of comparable closed models. That cost advantage gives Pinterest room to expand AI capabilities without the infrastructure expense that is constraining other platforms.

On the advertiser side, Performance+ delivered a 28% improvement in return on ad spend during testing for small and mid-sized advertisers using ROAS bidding. 

Smart Assembly, a new Performance+ feature that lets advertisers without product catalogues upload images for AI to automatically build and serve the best-performing ad, delivered a 6% average improvement in click-through rate in early testing. 

Internal AI tools for software development lifted weekly pull requests per engineer 45% year over year in July without a corresponding rise in incident rates.

Pinterest guided Q3 2026 revenue of $1.19 billion to $1.21 billion, 13% to 15% growth year over year, and raised its full-year adjusted EBITDA margin outlook to approximately 30%.