Shutterstock uses a custom-built neural network to power visual similarity search and reverse image search across its entire library, allowing users to find matching content by uploading a photo rather than typing keywords. This applies to both images and video.

Shutterstock launched reverse image search and visually similar search in March 2016, powered by a custom-built convolutional neural network that analyzes pixel data rather than metadata. The system has been continuously expanded: it now covers images, video, and is exposed via a developer API. Shutterstock later extended the neural network to video, allowing users to upload a photo or frame and receive video clips matching its visual composition, light temperature, and subject matter.