Details
- Perplexity AI has launched its public Search API, providing developers with direct access to an internally developed index covering billions of webpages.
- The service delivers real-time results with a median latency of 358 ms and keeps latency below 800 ms for 95% of queries, outpacing competitors by about 150 ms in internal tests.
- The product is structured around three key pillars: fresh and comprehensive data delivered quickly, advanced content understanding, and a hybrid retrieval system with multi-stage ranking and reranking.
- A distributed crawler processes approximately 200 million queries daily, while large language models break down and analyze pages, feeding a continuous loop that improves snippet accuracy.
- Perplexity has also released an SDK and an open evaluation framework for benchmarking search APIs tailored to AI agents; its API led in both quality and speed on single-step and deep-research assessments.
- The move follows Microsoft’s retirement of the Bing Search APIs in August, positioning Perplexity as a timely replacement for developers seeking dependable, real-time web data.
Impact
This launch serves developers affected by the discontinuation of Bing’s API and raises the stakes for competitors such as Google Programmable Search, You.com, and DuckDuckGo. By offering faster, more neutral access and handling its own crawling stack, Perplexity could accelerate AI-agent adoption and position itself as a strong challenger in consumer search, especially as Google faces legal and technological scrutiny.