Read the citation list as a SERP
For each target prompt, log the cited URLs in order. That ordered list is effectively a generative SERP, and the patterns across it - page type, publish date, presence of tables - tell you the format cost of entry for that query.
Reverse-engineer the winning passage
Open each cited page and find the sentence the answer paraphrased. Nine times out of ten it is a short declarative statement containing a number. Write your equivalent statement, with your own number, and put it high in the relevant section.
Publish the artifacts Perplexity likes to cite
Comparison tables, pricing pages, dated benchmarks, definition pages, and stat roundups are disproportionately represented in citation lists because they resolve sub-questions cleanly. Narrative thought-leadership posts almost never appear.
Keep pages crawlable to PerplexityBot
Perplexity's crawler needs access, and it strongly prefers pages that render content server-side. A client-rendered page that needs JavaScript to show its main text is effectively invisible to the fetch step.
Use follow-up questions as a keyword tool
The related questions Perplexity suggests are a live map of the sub-intents attached to your topic. Turn each into a question-shaped H2 with a standalone answer, and one page can start earning citations across the whole cluster.
Track share of citations weekly
Count, per prompt, how many of the cited sources are yours. Rising citation share is a leading indicator for referral traffic from AI assistants, which typically arrives weeks after the citations do.