The tools: ChatGPT, Perplexity, Claude — which for what

Same logic as the model-comparison course — pointed at research this time.

Perplexity: speed plus sources

Every answer comes with footnotes by default, search is quick, and the deep-research mode runs a few minutes. Reach for it when: you need a fast fact-check, a two-minute "what's known about X", a first look at an unfamiliar topic. The free tier is generous.

ChatGPT deep research: thoroughness

It works longer (10–20 minutes), reads more, and its reports are long and well structured; it holds up under complex multi-part requests. Reach for it when: this is the study your decision rests on — exactly the request format from the last lesson.

Claude: documents plus reasoning

Its strength is going deep on material you supply: drop in five competitor reports, an official statistics release, a long PDF — then ask for the synthesis. Web search is there too for freshness. Reach for it when: half the material is already sitting on your drive.

The researcher's combo

1. Scout (Perplexity, 5 minutes): what is this topic, what are the terms, where's the argument.
2. Go deep (ChatGPT DR, 20 minutes): the main report, built on the request formula.
3. Bring your own (Claude): your documents plus the findings → synthesis and comparison.

For small jobs one tool is plenty. For decisions where being wrong costs you, the combo pays for itself: different systems surface different things, and where they disagree is the most informative part of all (module 2 covers it).