NTLSN · Crash Course · AI skills

Prompting for academics — a crash course

AI is only as good as the brief you give it. Four short lessons on prompting well for academic work, then a self-check.

The one thing to remember: a good prompt is a good brief to a sharp research assistant: who to be, what to do, the context, and what ‘good’ looks like. Vague in, vague out.
4 lessons~11 min read1 self-checkGrounded in prompt-engineering practice (incl. Anthropic's guidance)

The lessons

1
The anatomy of a good promptRole · task · context · format

Most weak outputs come from weak prompts. A strong one sets a role, a clear task, the context, and the format you want back.

  • Give it a role (‘act as a curriculum designer…’).
  • State the task precisely, with the audience and purpose.
  • Provide the context and constraints; show an example of ‘good’.
  • Ask for the format you want — a table, five bullets, a rubric.
2
Iterate, don't expect one shotIt's a conversation

The first answer is a draft. The skill is steering — telling it what to keep, change, or go deeper on.

  • Refine: ‘shorter’, ‘more critical’, ‘with examples from nursing’.
  • Ask it to critique its own answer, then improve it.
  • Keep your best prompts — build a personal library.
3
Context is kingGive it your materials

AI can't read your mind or your course. Paste in the rubric, the brief, the reading — grounded prompts give grounded answers.

  • Paste the actual source (your outline, policy, draft) into the prompt.
  • Tell it what NOT to do, and the boundaries.
  • Ask it to work only from what you provided.
4
Verify everythingConfident is not correct

AI is fluent and sometimes wrong — including invented facts and citations. You own everything you use.

  • Check every fact and citation against a real source.
  • Never paste confidential or identifiable student data.
  • Disclose AI use per your institution's and publishers' policies.
Grounded in
  • Anthropic prompt-engineering guidance
  • Academic-integrity & AI-disclosure policy
◇ Bring it together — from the NTLSN commons

Before your next AI prompt — a quick self-check

My prompt sets a role, task, context and format.
I gave it an example of what ‘good’ looks like.
I treat the first answer as a draft and iterate.
I pasted in the real materials, not a vague description.
I verify every fact and citation against a real source.
I never paste identifiable student data, and I disclose AI use.
Source & attribution. Curated from prompt-engineering good practice, including Anthropic's published prompting guidance, indexed alongside the NTLSN commons. NTLSN is independent and not affiliated with Anthropic or any AI vendor — tools are named as widely-available examples. Always follow your institution's AI policy.
A free NTLSN crash course · see them all at /crash-courses.html. Runs in your browser; nothing is stored or sent.  ·  ← Back to NTLSN