Worked Humanities Workflows¶
These workflows show how to use ChatGPT's tools for common humanities tasks. Each suggests a starting point and the right tool for the job. For verification practices that apply across all platforms, see Verification & Citation.
1. Article or chapter drafting¶
Tool: Ordinary chat or a project. Canvas for heavily iterative writing.
Upload your abstract, notes, or draft section. Ask ChatGPT to identify argument structure, missing transitions, duplicated claims, implicit assumptions, and likely reader objections. Move into canvas if the writing becomes heavily iterative. Use Temporary Chat if you do not want the work to affect memory.
2. Current-context briefing¶
Tool: Web search or deep research.
Ask for a tightly bounded question, a short report, and explicit links back to sources. This is useful for policy changes, funding landscapes, software changes, or public debates. It is less suitable for settled historical scholarship or close bibliographic work.
3. Working with PDFs, spreadsheets, or catalogues¶
Tool: Data analysis (upload the file directly).
Ask for summaries, patterns, anomalies, missing values, column explanations, or quick charts. Treat this as exploratory analysis, not as a final statistical workflow.
4. Teaching preparation¶
Tool: Ordinary chat. Study mode for interactive learning.
Use ordinary chat for seminar questions, glosses, activities, assignment prompts, or alternative explanations. Use Study mode when the goal is learning through interaction rather than receiving a single answer. This can be helpful both for students and for staff learning a new method or technology.
5. Long-running research themes¶
Tool: Projects.
Use Projects when the work has continuity across sessions: a grant bid, module, edition, dataset, or book project. Add key files, keep task-specific instructions in the project, and avoid mixing unrelated work in the same project unless you want the context to blur.
6. Repeated specialist tasks¶
Tool: Custom GPTs.
Once a repeated workflow is stable, consider a custom GPT. This is sensible only when the task recurs often enough to justify the setup: for example, a style guide assistant, metadata normaliser, or recurring feedback assistant. Otherwise, the overhead is not worth it.
7. Technical build work¶
Tool: Codex.
If you are building software, websites, pipelines, or scripts, ordinary ChatGPT may help at first, but Codex is the more relevant OpenAI tool for real coding work. If you are not building software, skip it.