The Topics API aims to facilitate interest-based advertising by exposing coarse-grained topics derived from a user's browsing history. This API addresses the need for personalized advertising in a privacy-conscious environment. It leverages browser-side inference to determine user interests without requiring cross-site tracking. The primary technology involves a classifier model that maps visited hostnames to topics.
The API supports both JavaScript and request header-based retrieval, offering flexibility in implementation. It incorporates a taxonomy to categorize user interests, aiming to exclude sensitive topics. The design considers infrequent and frequent site visits to ensure sufficient topic coverage. It also includes a header-based alternative which improves performance for many use cases.
- Topic Inference: Browser-based classification of browsing history to determine user interests.
- Privacy-Preserving: Operates without cross-site tracking, enhancing user privacy.
- Flexible Retrieval: Supports both JavaScript API and request header-based retrieval.
- Taxonomy-Driven: Leverages a taxonomy of topics to categorize user interests.
- Epoch-Based: Returns topics in weekly epochs for consistent interest tracking.
- Header Support: Request headers provide a performant alternative to the JavaScript API.
- Deterministic Randomness: Uses a pseudorandom function to select topics.
The Topics API is currently archived and no longer being developed, as it has been deprecated by Chrome. The API's specifications and implementation were undergoing standardization and iteration, with ongoing refinement based on industry feedback. The project saw active development, including taxonomy updates and classifier model improvements, but maintenance has ceased.
The Topics API offered a solution for interest-based advertising by inferring user interests from browsing history. It provided a privacy-focused method for delivering relevant ads to users while addressing the limitations of third-party cookies. While now deprecated, the project demonstrated an approach to personalized advertising that prioritizes user privacy and site-level control.