If I say don’t use AI, you’ll still do. Hence, it’s better I teach you how to use it reasonably and properly.
If you want AI to help you choose a research topic, I would start with the circumstances of your project. As in, what are you studying? What interests you? What evidence can you get? All those details have to be in your prompt prompt before you ask for suggestions.
What I'll tell you is to treat the prompt as a short research brief. Give the AI something concrete to work with, ask it to explain the demands of each suggestion, and then question its recommendations.
I'm giving you some guide on how to do that in this article, with a template you can copy and a worked example of turning a broad interest into a manageable question.
I will also give you some sample prompt templates to copy and use for yourself.
Ask for a research question, not only a title
Consider this request for example:
Give me five unique research topics about social media and education.
The prompt you are seeing above says nothing about the course, the project requirements, the people you could reach, or the evidence you could use. I would not use it to choose a project.
There is also a difference between naming an area and identifying something to investigate. George Washington University's guidance on developing a research question explains that after choosing a topic, you need a specific, focused question.
For example, “social media and learning” names an area. “How do first-year students describe using course-related WhatsApp groups to clarify assignment instructions?” gives you something more definite to examine. This is an illustrative question, not a claim that the study is original or suitable for every programme.
When prompting AI, ask for the question alongside the working title. A polished title should not be allowed to hide an unclear study.
Give the AI the details that will shape your choice
Before writing the prompt, make a few notes about your project. Johns Hopkins University's topic-development guide asks researchers to consider their existing knowledge, disciplinary perspective, timing, and project length. I would bring those considerations directly into the conversation with AI.
Your subject and academic level: include your programme and whether this is a course paper, dissertation, thesis, or another kind of study.
Your actual interests: name a problem, debate, text, process, or experience you want to understand. “Business” gives very little direction; “how small retailers handle late customer payments” gives much more.
The requirements: explain any prescribed methods, topic boundaries, word limits, or expectations about an original contribution.
Your access to evidence: say whether you can use published literature, public datasets, archives, equipment, organisational records, or potential participants. Distinguish possible access from permission already obtained.
Your limits: include the deadline, budget, relevant skills, and any practical restrictions.
Be honest about what you do not know. If you have no preferred method, say so. If you cannot collect primary data, make that explicit rather than asking the AI to design interviews you will never conduct.
For a literature-based project, your evidence might be journal articles, historical documents, novels, policy texts, or philosophical arguments. You do not need to invent a participant group simply because the template includes a space for one.
Copy this research-topic prompt
Replace the bracketed sections with your details. Where something does not apply, write “not applicable.” Where you are uncertain, explain the uncertainty.
Help me develop a focused research topic that fits my project requirements and the resources available to me.
My subject or discipline: [subject].
My academic level and project type: [level and type of assignment or research project].
My interests: [two or three specific issues, questions, debates, or problems].
Relevant setting, population, materials, or period: [details, if applicable].
My deadline and project length: [time available and word or page limit].
Evidence or resources I can access: [literature, datasets, texts, archives, equipment, records, or potential participants; explain whether access is confirmed].
My skills and limitations: [methods experience, budget, access restrictions, or other constraints].
My department's requirements: [requirements, or explain what still needs confirming].First, ask up to five clarifying questions if important details are missing. Wait for my answers before suggesting topics. If you have enough information, proceed.
Suggest three distinct topic options. For each option, provide:
A clear working title.
One main research question.
The problem or uncertainty the study would examine.
A possible approach or method, with a brief explanation of why it fits the question.
The evidence needed and any access or permission issues.
The main feasibility concern and one way to narrow the scope.
Search terms I can use to investigate the existing literature.
Compare the options against my requirements and explain which seems most workable. Make assumptions explicit. Do not invent sources or claim that a topic is original. Treat any suggested research gap as provisional until checked against the literature. If you cannot verify a factual claim or reference, say so.
The useful part of this template is the brief you supply and the explanation you request for each option. You can shorten it when you already know your requirements. Keep the evidence and feasibility questions: those are the parts I would want answered before choosing a topic.
See how the brief changes the task
Let us stay with social media and learning. Here is a more useful request than asking for “unique topics.” This is a hypothetical teaching example.
I am preparing an undergraduate education project and have twelve weeks to complete it. I am interested in how students use messaging groups for coursework. I may be able to recruit adult first-year students at my institution, subject to permission and ethics requirements, but I cannot access their grades or private chat histories. Suggest three focused questions that do not depend on those records. Explain the evidence each question would require and the main limitation.
Notice what the student has done. They have identified an interest, described a possible route to evidence, and ruled out records they cannot access. That gives us a basis for judging suggestions.
One candidate to examine could be:
How do first-year students describe using course-related WhatsApp groups to clarify assignment instructions?
For this example, I would explore interviews about students' experiences instead of propose an analysis of private messages. I would keep the question about those experiences and not turn it into a claim about improved grades. The proposed question still needs checking: does it meet the department's expectations, can participants be recruited appropriately, and does the literature provide a worthwhile reason to investigate it?
This is where feasibility becomes concrete. The University of Southern California's research-problem guide connects researchability with access to information that can actually be gathered and analysed. Apply that test to every AI suggestion, however attractive the title sounds.
Use a follow-up prompt to challenge the strongest option
Once you have three options, I would resist asking for another twenty. Choose the most promising one and examine its weaknesses.
Use this follow-up to make the next exchange more useful:
Take option [number] and assess it critically against my project brief. Identify vague terms, hidden assumptions, evidence I may not be able to obtain, and any mismatch between the question and the proposed method. Give me a narrower version and explain what I would lose by narrowing it. Do not add complexity unless it is necessary to answer the question.
Read the response as a set of issues to investigate, and not accepting its verdict. For the messaging-group example, I would ask what “using” includes: reading announcements, asking questions, sharing files, or discussing instructions. I would also ask whether focusing on one course would make the project more coherent or leave too little to investigate.
The aim is a question you can explain plainly. If you still cannot say what evidence would answer it, keep working on the question.