Perplexity and the answer engines

Ask a librarian a question and they don’t recite from memory — they walk to the shelves, pull three books, and answer with the pages open in front of you. That’s the answer-engine idea, and Perplexity made it a product category.

The fact sheet (community-verified)

Search-first, cite-always

A chat assistant defaults to pattern memory and reaches for search as a tool. An answer engine inverts this: every question triggers retrieval first, and the model composes its answer from the retrieved pages, citing them inline. Two practical consequences:

Recency by default. Questions about current events, prices, releases, and “what’s the latest on X” play to its core design rather than an added feature.
Verifiability by default. Citations arrive without asking — which moves you instantly to the strong position from L4: checking sources rather than bare claims.

Where it fits — and where it doesn’t

Reach for an answer engine when the task is finding out: research, comparisons grounded in current information, “is this still true?” checks — including checking claims other assistants made. Reach for a chat assistant when the task is making: drafting, rewriting, brainstorming, transforming your own material — work where retrieval adds little and conversation depth matters.

Reading the citations well is the skill: notice which sources were retrieved (a vendor’s press release and an independent review support claims differently), notice when several citations trace back to one original, and remember the composed summary can still stitch sources together imperfectly — the sentence-to-source link is yours to confirm for anything that matters.

Try it now (4 minutes)

Take one factual claim an assistant gave you earlier in this module and put it to an answer engine: “Is it accurate that …? Cite sources.” Open the two strongest citations. Confirmed, corrected, or complicated? All three outcomes are wins — you now know something with sources attached.

Check your understanding

1. Which task most favours an answer engine over a chat assistant? Rewriting your notes into a friendly email Finding what changed in visa rules this year, with sources Brainstorming names for a bakery

2. An answer arrives with five citations. The remaining verification step is: None — five citations settle it Confirm the key sources actually say what the answer claims Count whether other tools give more citations

Recap

Answer engines search first and cite always — strongest for finding out, not for making; citations move verification to sources, not remove it. Next lesson: Copilot and assistants living inside your working documents.