Perplexity's Decisions API can now search the web before answering, and on Perplexity's own test true/false accuracy rose from 56.4% without search to 97.3% with it.

A model answering from training knowledge alone gets facts that change over time wrong, and searching first fixes that at 2.6 times the median latency. The test used 328 questions Perplexity selected, mixing October news and obscure facts. The Decisions API is a service that lets a program ask yes/no, multiple-choice or scoring questions and returns probabilities. Perplexity writes that accuracy is 56.4% without search and 97.3% with it, and median latency goes from 139 milliseconds to 366 milliseconds. On PWS-Decisions, a public 100-question benchmark, accuracy rose from 51% to 86%. Each search costs an extra US$0.001 (Perplexity, 2026-11-04). How we checked →
Source and checks
This item ran in the brief of Nov 5, 2026, alongside 13 other items. How we checked it, the quote checks and what was still missing that day are on that issue's How we checked page.
Cite this item
SecondSource (2026). "Perplexity's Decisions API can now search the web before answering, and on Perplexity's own test true/false accuracy rose from 56.4% without search to 97.3% with it.." https://secondsource.io/en/product/2026-11-05-2/. Accessed access date.