Capstone: a real research task, end to end
Goal. Combine everything — search, files, analysis, and verification — to take one real question from background reading to an analyzed result with checked sources.
I'm investigating whether [gene X] expression differs between [condition A] and [condition B]. Step 1: search the web for 3 recent primary papers on this and summarize each with a link. Step 2: I'll upload my expression CSV — run the appropriate test and plot it. Step 3: draft a 120-word results paragraph citing only my data, and list which of the searched papers I should verify before citing.
Run it as three steps in one chat. This is where Deep research helps for the literature step — a longer, multi-source search that returns a cited report.
- 1Background (search + verify). Have ChatGPT search and summarize recent primary literature with links, then open each link and confirm it before trusting it (Lesson 3).
- 2Analyze (your files). Upload your data and have ChatGPT run the correct test and produce a figure (Lesson 2), checking the stats yourself.
- 3Write and check. Have it draft a results paragraph from your data only, and flag every external citation for you to verify. For deeper literature reviews, try Deep research, which runs an extended multi-source search and returns a longer cited report.
You'll see. A complete mini-workflow: verified background sources, an analysis and figure from your own data, and a drafted paragraph with a clear list of citations still to check.
Cost. The core workflow runs on the free tier (with limits on uploads, analysis, and search). Deep research is limited on free and Go, and expanded on Plus and Pro — reach for a paid tier when you do serious literature work.
Takeaway. ChatGPT's real power for research is the combined loop — search, upload, analyze, draft — with a human verification step at every stage you can't skip.