E040 pre-registration: query surface form and citation rate on Perplexity (keyword, interrogative and conversational forms against three fixed pages)
Abstract
This pre-registration sets out an observational study of whether changing the wording of a query affects whether Perplexity cites a page, while keeping the concept and target page fixed. The study uses the direct Perplexity API with the model string `sonar`, temperature 0 and `search_context_size` set to low. The instrument was pinned on 2 October 2026. The design covers three pages, with five concept clusters per page and three versions of each query: keyword, interrogative and conversational. That gives 45 queries and 225 measurements across five days. The primary comparison is between the keyword and conversational forms on the two clean pages, `/extractability/` and `/retrieval-probability/`. The `/geo-stack/` page is analysed separately because it is under a content freeze and has a different citation baseline. The frozen query set, SHA-256 `e9415f33ab28e676e4613bf689330a0c3c84d9d9610865042f81eeffe61a8571`, and the instrument pin are included. No measurements were taken before registration.
// Source
Authors: Artur Ferreira