How to Choose a Thesis Topic (When Nothing Feels Good Enough)

The short answer

Choose a thesis topic by starting from a question rather than a subject, then testing that question four ways: is it answerable with the time and access you have, do enough sources already exist, is it narrow enough to answer within your word count, and can you stay interested in it for months. Most candidate topics fail the source test or the scope test, and both can be checked in an afternoon — before you commit a semester to finding out the hard way. The topic does not need to be original at bachelor's or master's level; it needs to be finishable.

Why you are stuck (it is usually not lack of ideas)

Students who cannot choose a topic almost never have zero ideas. They have several, and no way to compare them, so they wait for one to start feeling obviously right. That feeling does not arrive, because it is not produced by thinking — it is produced by testing.

There are three failure patterns behind most of it:

• Waiting for originality. You are looking for something nobody has done, which at undergraduate level is the wrong target entirely. See the FAQ below — this is the single most common misunderstanding about what a thesis is for.
• Choosing a subject, not a question.“Something about sustainability in supply chains” is not a topic. It cannot be answered, so it cannot be finished. A topic is a sentence ending in a question mark.
• Skipping feasibility. The idea is genuinely interesting and the data does not exist, is behind a paywall your library does not have, or would require access to a company that will not grant it. This is discovered in month three rather than week one.

The four tests a topic has to pass

Run every candidate through all four. Write the answers down — the point is to compare candidates on the same criteria instead of by how each one feels on the day.

1. Answerable. Can this be settled with the method, time and access you actually have? A question needing three years of longitudinal data fails in a semester.
2. Sourced. Do at least 15–20 relevant, accessible academic sources already exist? If almost nothing exists, that is far more often a signal the question is unworkable than a gap nobody noticed.
3. Scoped. Could you answer it properly in your word count? If it would need three chapters just to define terms, it is too broad.
4. Sustainable. Can you read about this for months without resenting it? This test is last for a reason — it only matters among topics that already passed the first three.

How to narrow a topic that is too broad

Nearly every first draft of a topic is too broad, and narrowing is mechanical rather than creative. Take the broad version and add constraints one at a time:

• Population:which people, firms, or cases — “German SMEs” rather than “companies”.
• Period:which years — “since 2020” rather than “recently”.
• Context: which sector, region, or setting.
• Relationship: which specific effect on which specific outcome.

Worked example. Start with “remote work and productivity” — a subject, not a question. Add a relationship: “How does remote work affect productivity?” — answerable in principle, far too broad in practice. Add population and context: “How do hybrid work arrangements affect self-reported productivity among software developers in German SMEs?” Add a period and you have a thesis question that can be answered in 8,000 words and whose literature you can actually read in the time available.

If narrowing feels like you are giving up the interesting part, that is normal and it is wrong. The narrow question is the one you can answer well, and an answered narrow question marks higher than a gestured-at broad one.

The 30-minute source check that saves a semester

This is the highest-value half hour in the entire process, and most students skip it. Before committing to a topic, do this:

1. Search your question's key terms in Google Scholar and in your university library database. Not Google — the databases you will actually have to cite.
2. Count how many accessible results are relevant and from the last ten years. If you cannot reach 15 without stretching the definition, treat the topic as failed.
3. Open the three most relevant papers and read their “limitations” and “future research” sections. This is where supervisors' favourite thesis topics come from, stated explicitly by researchers in the field.
4. Check one paper's reference list for names that keep recurring. If the same five authors appear everywhere, you have found your core literature and the topic is viable.

Step 3 deserves emphasis. Published papers routinely end by naming what they could not do — a population they did not cover, a period that has since passed, a method they did not apply. Those sentences are pre-validated thesis topics: a researcher has already argued the gap matters, in print, which is most of your justification section written for you.

What changes by level: bachelor, master, PhD

The four tests are the same. What changes is how much originality is expected, and this is where most anxiety about “is my topic good enough” comes from.

• Bachelor's. Demonstrate that you can apply a method correctly and engage with existing literature. Applying a known framework to a new context is entirely normal and expected. Nobody requires a novel contribution.
• Master's.The same, with more methodological independence and a clearer statement of what your work adds. A replication with a different population or more recent data is a legitimate master's thesis.
• PhD. Here originality genuinely is the requirement, and topic selection is a different exercise conducted over months with a supervisor.

How to bring it to your supervisor

Do not arrive with “I was thinking about something in this area”. Arrive with two or three candidate questions, each with the four tests answered, and ask which is most feasible. This changes the conversation from them generating ideas to them evaluating yours, which is what supervisors are good at and what they expect.

Bring, for each candidate: the question in one sentence, roughly how many sources you found and where, the method you would use, and the single thing you are least sure about. That last item is what a supervisor can help with fastest.

Should you use AI to find a topic?

For generating candidates and narrowing a broad subject, it is genuinely useful — that is a brainstorming task, and it is exactly what students are already searching for. It is reasonable to ask a model to suggest twenty questions in your field, or to narrow one by population and period, and then run the survivors through the four tests yourself.

Two limits worth knowing before you rely on it. First, language models invent plausible citations — a suggested topic that comes with references is not evidence those references exist, and the source check above is not optional. Second, whether AI assistance is permitted at your institution varies, and topic selection is usually treated far more permissively than drafting text. Check your regulations rather than assume.

If you do end up using AI to help draft any part of the thesis itself, that is a different question with different rules, and it is worth knowing what a detector would make of your writing before your university runs one.