Exercise 5: Deep Research on Golden History
Most of the time, AI gives you a quick answer from memory. Deep research is different — it actually goes out and reads the web, works through dozens of sources, and comes back with a synthesized report complete with citations. The tradeoff: it takes time. A lot of it.
This exercise will take 15–30 minutes to complete — and that's normal.
Deep research isn't a faster version of a regular search. The AI is actively reading web pages, following links, cross-referencing sources, and writing a structured report as it goes. If it looks like nothing is happening, it's working. Don't close the tab.
When it finishes, you'll get a long, well-organized report with citations you can actually click. That's the point — the answer is grounded in real documents, not generated from memory.
Look for the deep research button in your tool — availability is expanding.
- ChatGPT — Click the "Deep Research" button in the chat interface
- Gemini — Click "Deep Research" (available on the free tier)
- Claude — Click the "Research" button — same idea, different name
- Copilot — Has research capabilities; look for a research or search toggle
Availability varies by plan and changes frequently. If you don't see the button, try a regular prompt — you'll still get a useful answer, just not a multi-source report.
The exercise
Pick a topic below and let AI do a deep dive into Golden's history. These are topics with rich, well-documented histories — good material for a research report.
Prompt
Do a deep research report on [topic]. Focus on [focus]. Present the results as [format]. Cite your sources throughout so I can read further.
What to look for when it finishes
The citations — Scroll to the end of the report and look for footnotes or source links. These tell you exactly where the AI found its information. Click a few to verify they're real and that the AI represented them accurately.
Confident claims worth checking — Deep research is more reliable than a regular prompt, but not infallible. If the AI states a specific date, name, or statistic, spot-check it against one of its own cited sources.
The depth — Notice how much more detailed the report is compared to a standard answer. This is what grounding in real documents looks like.
Try the same prompt on two tools and compare. ChatGPT and Gemini both have deep research modes. Running the same topic through both and comparing the results — what each found, what each emphasized, what each got slightly wrong — is itself a useful lesson in how these systems work.