Brainstorming Topics & Outlining Your Paper
The blank page is the enemy of every student. Brainstorming with AI lets you generate dozens of angles, narrow to one, and build a structured outline in a single 30-minute session. Done right, it dramatically reduces procrastination and helps you find topics you actually find interesting. Done wrong, it produces a generic outline that any other student could have generated — and that your professor has seen 50 times this term.
This lesson teaches the difference.
What You'll Learn
- How to brainstorm sharp, specific research questions instead of generic topics
- A two-stage outlining workflow that produces a defensible structure
- How to use AI to find angles other students will miss
- How to pressure-test your outline before you start writing
The Brainstorming Mindset
Most students approach brainstorming by asking the AI a single question: "Give me ideas for an essay on [topic]." The AI returns five generic angles. The student picks one. The result is a paper that looks like every other paper.
A better approach: treat brainstorming as a conversation that converges. You start broad, push for specificity, ask "what would someone disagree with here?", and iterate until you have a question that is both interesting and tractable.
A research question is good if it has four properties:
- Specific. Not "AI and education" but "Does the use of ChatGPT in first-year writing classes affect students' revision behavior?"
- Answerable. You can imagine what evidence would settle it.
- Contested or unsettled. Smart, informed people disagree about the answer.
- Within scope. You can address it in the word count and time you have.
AI is great at helping you find this kind of question, but only if you push it.
Stage 1: Broad Brainstorm
Open ChatGPT, Claude, or Gemini. Use this prompt:
Act as a senior researcher in [discipline]. I am a [year] [level] student writing a [word count] paper for a class on [course topic]. My broad interest is [your topic].
Generate 10 specific research questions on this topic, ranging from very narrow to broader. For each, briefly state:
- Why someone might find this interesting
- What kind of evidence would answer it
- One reason another smart person might disagree with the most common answer
Avoid generic, well-trodden questions. Push for angles that a reader would not immediately predict.
You will get a varied list. Some will be too narrow, some too broad, some genuinely interesting. Scan the list and circle three that catch your attention.
Now push back:
Of those, questions 2, 5, and 8 are most interesting. For each, suggest two ways to make the question more specific or more interesting. Then suggest one common pitfall a student writing on this question is likely to fall into.
After this exchange you will usually have a question that is sharper than any of your initial drafts.
Stage 2: Validate the Question
Before committing, validate that the question has a real literature to engage with. Run it through Elicit or Consensus:
"Does the use of ChatGPT in first-year writing classes affect students' revision behavior?"
If you get a healthy set of papers, the question is researchable. If you get only one or two papers, either you need a different question, or you have stumbled onto a genuine gap (which can be a strength, but is harder to write about as a beginner).
Also do a Google Scholar search. Sort by recency. If there is a substantial recent literature, you have a workable topic.
If there is nothing — pause. A "gap" with zero literature is usually a sign that the question is unanswerable, not that it is original.
Stage 3: Building the Outline
You have done some reading (literature review skills from Lesson 7), you have a thesis, and you have a question. Now you build the outline.
Use this prompt with Claude or ChatGPT:
Act as an academic writing coach. I am writing a [word count] paper for a [discipline] class. My research question is: [paste]. My working thesis is: [paste a one-sentence thesis]. The main sources I will draw on are:
- [Author (year): main claim, in one sentence]
- [Author (year): main claim, in one sentence]
- [continue for 6–12 sources]
Suggest a logical outline with 4–6 sections. For each section give:
- The heading
- Two to four bullet points of sub-arguments
- Which of my listed sources I would cite there
- The approximate word count for that section
Do not introduce sources I have not listed. Do not write the actual paper.
The output should be a structured outline that uses your sources, fits your word count, and follows a logical argument. If the outline introduces sources you did not list, the AI is hallucinating — discard those.
Stage 4: Pressure-Test the Outline
Before writing, stress-test the outline. Ask:
Here is my proposed outline: [paste]. Identify the three weakest parts of this outline:
- Sections where the logic does not clearly flow from one to the next
- Sections where I am likely to make a claim my sources cannot fully support
- Sections that are likely to feel like filler
For each weakness, suggest one concrete improvement.
This kind of red-team prompt makes your outline much stronger. It is a step most students skip.
You can also ask:
What is the single strongest objection a reviewer might raise against the argument structure I have outlined? Where in the paper should I address it?
A paper that has anticipated and addressed an objection is dramatically more sophisticated than one that has not.
Finding Non-Obvious Angles
The strongest student papers find an angle other students will miss. Here is a prompt that helps:
Most students writing on [topic] take one of the following angles: [list 3-4 common angles]. Suggest five angles that are less common but still defensible and grounded in real research. For each, explain why it might be more interesting than the common angles and what evidence would be needed.
Combine this with the "two reasons people disagree" prompt, and you will end up with topics most other students will not have considered.
A Common Trap: The Five-Paragraph Essay
A trap students fall into when using AI for outlining is the five-paragraph essay structure: introduction, three body paragraphs (one per source), conclusion. This is fine for a 500-word essay. It is wrong for a 2,000-word college paper, where each "body" should weave multiple sources into a thematic argument, not a list of summaries.
If your AI-generated outline has one paragraph per source, ask:
Restructure this outline so that each section is organized by argument or theme, not by source. The same source may appear in multiple sections; multiple sources may appear in each section.
The restructured outline will be a much more sophisticated piece of writing.
A Quick Exercise
Pick a topic from an upcoming or current assignment. Run Stage 1 (broad brainstorm with 10 angles) and Stage 2 (validate against Elicit). Pick one question. Then run Stage 4 (pressure-test) on a one-sentence draft thesis.
Total time: 30 minutes. Outcome: a sharp question, an outline, and a stress test — all done before you start writing.
Key Takeaways
- A good research question is specific, answerable, contested, and within scope. Push the AI to help you find this kind of question, not generic topics.
- Use a two-stage brainstorm: broad first, then narrow with iteration and validation against real literature.
- Outline by argument and theme, not by source. Avoid the five-paragraph essay trap.
- Always pressure-test your outline before writing. Red-team prompts surface weaknesses you would otherwise discover during writing — when fixing them is much more painful.
- Use AI to find non-obvious angles other students will miss, then validate that those angles have real literature behind them.