Artificial Intelligence (AI) is a type of computer science. It develops systems that can assist with tasks that may require human-like intelligence, such as problem-solving and understanding language. There are two forms of classical AI:
At Pinterest, we’ve always used both forms of classical AI to help people find inspiration to create a life they love. This can show up in a variety of ways, including:
Where appropriate, we use human reviewers and feedback from the public to refine our AI. We continuously monitor and evaluate our methods to work to ensure the experiences are working as intended.
Machine learning involves the development of algorithms and statistical models. It helps computers learn from and make decisions based on existing data. Machine learning systems are built to perform specific tasks, but are not manually programmed to do so. Instead, these systems find patterns, make predictions and can improve performance over time as more data is gathered.
Traditional machine learning is primarily about predicting or classifying existing data. For example, to teach a computer to recognize chairs in photos using machine learning, you'd show it a lot of pictures of chairs, and items that aren’t chairs. Then, the computer will learn by itself to tell the difference.
At Pinterest, we use machine learning in many different areas of the business. For example, we use it to recommend and show the most relevant ideas across home feed, search and related feeds. We also use it to increase inclusive representation across Pinterest’s recommendation feeds, such as diverse body types.
We’re thoughtfully exploring Generative AI (or Gen AI) technology that drives innovation and creativity. For example, we use Gen AI technology to make ads more engaging and improve campaign performance for advertisers by
We have taken steps to help ensure the responsible use of Gen AI. Our
We’re always making efforts to enhance our content moderation systems. As part of this, we have partnered with the Digital Trust and Safety Partnership (DTSP) and participated in their Gen AI and Automations Working Group.
Our
The labeling of AI-generated or modified content helps to provide relevant context about the content people see on Pinterest.
Pinterest may show a label on the close-up of an image Pin when we detect that it was generated or altered with AI. This is according to the
For ads on Pinterest, we provide a disclosure only in “Why am I seeing this ad?” showing whether the ad contains content that was modified or generated with AI, as detected through metadata.
We’re working on ways to expand capabilities to better identify Gen AI content in the future through additional technologies.
Learn more about
Pinterest’s Navigator-1 (“Navigator-1”) is a multimodal foundation model that ingests text transcribed from audio inputs to power Pinterest Assistant text outputs, for the purpose of supporting conversational experiences on Pinterest. Navigator-1 was built by fine-tuning a publicly available open-source base language model in January 2026, using synthetic data, and it will be periodically retrained using additional synthetic data. Pinterest used approximately 150,000 synthetically generated training text examples that simulate realistic multiturn conversations. The synthetic data set was rated, cleaned, and processed for quality purposes.
Because the conversation data set used to train the model was synthetically produced by Pinterest, it was not collected from users, purchased or licensed from third parties, and does not include personal information or aggregate consumer information. Further, it does not include third-party materials protected by copyright, trademark, or patent.
Pinterest Canvas Version 6 (“Pinterest Canvas”) is a multimodal foundation model that powers Pinterest experiences by enabling the enhancement of images. Pinterest Canvas has been trained on approximately 500,000,000 rows of publicly available Pins, Pin descriptions, and Pin metadata that Pinterest users have saved or uploaded to Pinterest, to which the users have granted Pinterest a license, as well as the continuous use of synthetic data. Training data rows may include one or more images together with text instructions and a target image. The data in the datasets were collected by Pinterest as early as 2009 and were first used in the development of Pinterest Canvas in or around January 2023. The data collection is ongoing. The data in the datasets may include limited personal information, and material that may be protected by copyright, trademark, and/or patent. The datasets do not include aggregate user information. The datasets were cleansed, processed, and modified for quality purposes and to meet our trust, safety, and other internal standards.
To