Perplexity's citations teardown

Claims carry numbered citation markers inline. Readers scan the answer first; numbers signal “this is sourced” without footnotes that interrupt rhythm.

Numbers also give users a shared reference (“see source 3”) when sharing or discussing the answer.

Use inline markers for search answers; burying links only at the bottom trains users to distrust the prose.

A single expander opens the full source list for users who want audit-level detail. Default view stays clean; depth is one click away.

List view supports skimming domains and titles before opening any link. That reduces tab spam for research workflows.

Pair inline numbers with an exhaustive list so quick readers and auditors both get what they need.

Drilling into one citation opens a card with context: where the source came from, what passage supported the claim, and quality hints where available.

Cards keep users in-product for verification; outbound links remain, but preview builds confidence first.

Show enough snippet to decide if the source is worth opening externally.

Read answer → spot numbers → expand all sources or open one card → follow out if needed.

Perplexity invests in output UX, not post-answer refinement menus. Iteration is usually a new query with Focus, not “make shorter” on the same bubble.

Steal the three-layer cite model if your product makes factual claims: inline, list, detail.