Perplexity: Search and generation in one surface
Perplexity rebuilt search around an LLM — and its UI earns trust through visible tool calls, sentence-level citations and a persistent sources rail.
An independent reading of Perplexity's interface, grounded in Perplexity's official documentation. Quotes are verbatim from the linked official pages. The product UI may differ from what is described.
Search as the default mode
- Every query is a search; the tool call is shown as a small visible step, demystifying where the answer comes from.
- Elapsed time and result counts signal real work happening — not a magic answer.
“Simply ask any question, and it searches the web to deliver accessible, conversational answers backed by verifiable sources.”
Source: Perplexity Help Center — What is Perplexity? ↗
Sentence-level citations
- Numbers attach to the claim, not the paragraph — a reader can verify the exact sentence.
- Hovering brings up the source card with domain, favicon and a snippet; credibility reads instantly.
“Each answer includes numbered citations linking to the original sources, allowing you to easily verify the information or explore further.”
Source: Perplexity Help Center — How does Perplexity work? ↗
The most capable models ship in two variants — one with additional safety measures, and one restricted to approved organizations. When rendering tokens incrementally, treat the stream as a first-class UX concern rather than a progress hack .
For actions with real-world consequences, a human-in-the-loop gate is not optional — it is the difference between a helpful assistant and an uncontrolled agent .
The sources rail
- A persistent sidebar lists every source used, ranked by relevance, with expandable snippets.
- Sources persist across follow-ups, so later claims stay checkable against the same thread.
“Select or hover over a source in an answer to open its details.”
Source: Perplexity Help Center — Understanding source labels ↗
Claude 5: A new era of AI capability
anthropic.com
Claude Fable 5 and Mythos 5 share the same underlying model, with Fable 5 adding safety measures for dual-use capabilities.
State of the AI UI stack in 2026
example.com
A survey of how teams are composing prompt inputs, agent states and streaming UIs across frameworks.
Designing human-in-the-loop approvals
example.com
Why approval requests need context, cost preview and an escape hatch before an agent acts.
Follow-up flow
- Suggested follow-ups keep the session moving; the thread reads like a conversation, not a query log.
- The composer is a single line — low friction, high volume — encouraging the next question.
“You can ask follow-up questions, and Perplexity will remember the context of your previous queries, ensuring a seamless, flowing conversation.”
Source: Perplexity Help Center — How does Perplexity work? ↗
Design takeaways
- Make the grounding visible: users trust what they can check, so show the search step.
- Attach citations to sentences, and keep the source card one hover away.
- Keep sources persistent across the whole thread — trust compounds turn by turn.
- A single-line composer with suggested follow-ups is the lowest-friction loop in AI UI.