Recorded prompt
“How should a new website measure citations in ChatGPT and Perplexity?”
- Product
- Perplexity
- Surface
- Perplexity Search with web search enabled
- Model
- Perplexity automatic model (model version not exposed)
- Locale
- en-US (English answer; zh-CN interface)
- Observed answer
- Open session ↗
- Observed
- 15 November 2026, 18:10 CST
- Fresh conversation
- Yes
What the answer said
For a new site, Perplexity recommended a fixed set of 20–30 customer questions grouped by unbranded discovery, long-tail questions and branded evaluation. It advised fresh conversations with consistent mode, language and location, followed by three runs per prompt and a recorded answer, mention, direct citation, exact cited URL, competitor and accuracy fields. It defined citation rate as answers linking to the domain divided by valid tested answers and distinguished that from brand mentions, distinct cited pages, AI referral sessions and referral conversions. It also said crawler activity is diagnostic rather than proof of citation and recommended preserving a baseline before changing the prompt panel.
10 recorded sources
What this observation cannot show
Authenticated free-account run in a fresh Perplexity conversation. Perplexity did not expose its automatic model version or retrieval index; the answer and source set can vary by session, account state and date.