Your First Research Prompts That Actually Work

A prompt is the instruction you give an AI model. Most students get bad results from AI not because the AI is bad, but because the prompt was vague. "Write me an essay on climate change" gets you generic mush. "Suggest three angles a sociology student might take on the relationship between climate anxiety and intergenerational political engagement, with one possible counter-argument for each" gets you something genuinely useful.

This lesson teaches you the prompting patterns that work specifically for academic research. By the end you will have a toolkit of templates you can copy, paste, and adapt for the rest of your studies.

What You'll Learn

• The four-part structure of an effective research prompt
• Six copy-paste prompt templates for common student research tasks
• How to push back when the AI gives a weak answer
• The single biggest prompting mistake students make

The Four-Part Prompt Structure

Every good research prompt has four pieces. You can remember it as R-T-C-F: Role, Task, Context, Format.

Role. Who is the AI pretending to be? "Act as a research librarian." "You are a graduate-level sociology tutor." "Act as a peer-review reviewer for a journal in molecular biology." Giving a role focuses the model on the right vocabulary and level.

Task. What exactly do you want? Be specific. Not "help with my paper" but "suggest three sub-questions I could explore" or "summarize this section in 150 words."

Context. What does the AI need to know? Your level of study, the discipline, the assignment requirements, the word count, the citation style, what you have already done.

Format. How should the answer look? Bullet points? A table? A 300-word paragraph? Headers and subheaders? Specifying the format saves you reformatting later.

Here is the difference in practice.

Vague prompt: "Tell me about renewable energy."

R-T-C-F prompt:

Act as an energy policy researcher. I am a second-year undergraduate writing a 1,500-word paper for an environmental policy class. Compare the policy frameworks for solar energy in Germany and the United States from 2015 to 2025, focusing on subsidies and grid integration. Give me three thematic angles I could focus on, with one example real research question for each. Format as a numbered list with one paragraph per angle.

The second prompt produces something usable. The first produces a Wikipedia article you could have written yourself.

Six Copy-Paste Templates

Copy these into a note-taking app and adapt them. Replace anything in [brackets].

1. Brainstorming research angles

Act as a [discipline] tutor at a research university. I am a [year] [level] student. My broad topic is [topic]. Suggest five distinct angles a paper on this could take, ranging from very narrow to broader. For each, give a one-sentence research question and identify what kind of evidence would support it.

2. Generating a working outline (after you have done the reading)

Act as an academic writing coach. I am writing a [word count] paper for a [discipline] class. My thesis is: [paste your draft thesis]. The main sources I plan to cite are: [list them]. Suggest a logical outline with headings and one bullet per sub-point. Do not invent sources I have not listed.

3. Explaining a confusing concept

Explain [concept] as if I were a [year] student in [discipline] who has read [what you have already read]. Use one concrete example and one analogy. Then list two common misunderstandings of this concept.

4. Pressure-testing your argument

I am arguing that [paste your thesis or claim]. Act as a skeptical reviewer in [discipline]. Give me three serious objections to this argument, with the strongest one last. For each, suggest what evidence would weaken or strengthen the objection.

5. Summarizing a paper you have already read

Below is the abstract and introduction of a paper. Summarize the authors' main argument in 150 words, list the methods they use, and identify the most important limitation they themselves acknowledge. Do not invent details. [paste text]

6. Suggesting search terms for a literature search

I am doing a literature review on [topic] in [discipline]. Suggest 10 specific search queries I could use in Semantic Scholar or Google Scholar, ranging from broad to narrow. Include relevant Boolean operators and quoted phrases. Note any common synonyms or alternative spellings I should also try.

Notice what is missing from all of these: requests for the AI to find sources or generate citations. We do that with academic tools (next lesson). General AI tools are for thinking, not finding.

How to Push Back on a Weak Answer

If the first answer is generic, you do not have to accept it. Continue the conversation.

• "That is too generic. Give me a version specific to [your discipline / region / time period]."
• "You repeated yourself in point 2 and point 4. Combine them and give me a new fifth point that introduces a different angle."
• "Make the language more academic — fewer rhetorical questions, more precise vocabulary."
• "Cut this from 400 words to 200, keeping the key claim."
• "What would a critic of this argument say?"

A conversation is more powerful than a single prompt. Treat the AI as a junior collaborator who needs direction.

The Biggest Mistake: Asking for Sources

The single most common mistake students make is asking the AI to find sources for them.

"List 10 peer-reviewed articles on [topic] published in the last five years, with DOIs."

The AI will produce a list. The list will look perfect: real-sounding author names, plausible journal titles, valid-looking DOIs. And often none of them will be real. The papers do not exist. The DOIs lead to 404 errors. The authors may exist but have never co-written that paper.

This is called hallucination, and it is the most common way students get caught using AI badly. A professor sees "Patel & Ramirez (2023)" in your bibliography, searches for it, finds nothing, and now you have a major problem.

The rule is simple: never use a citation that came from a general AI tool unless you have personally verified it in Semantic Scholar, Google Scholar, your library catalog, or the original journal.

Use general AI for thinking. Use academic AI (which we cover in the next module) for finding sources.

A Quick Exercise

Pick any topic you are currently studying. Open ChatGPT (or Claude or Gemini). Write a prompt using the R-T-C-F structure to get five research angles on the topic. Then iterate at least twice — refine, push back, ask for more specificity. Save the final answer.

Do the same prompt using only the vague version: "Give me five angles on [topic]." Compare the two.

This contrast is what good prompting feels like.

Key Takeaways

• Effective prompts have four parts: Role, Task, Context, Format (R-T-C-F).
• Use template prompts for brainstorming, outlining, explaining, pressure-testing, summarizing, and suggesting search terms — but adapt them.
• Treat prompting as a conversation. Push back, ask for revisions, demand specificity.
• Never ask general AI tools to find or generate citations — they will hallucinate. Use academic tools for finding sources, general tools for thinking.