There's a particular moment every journalism student hits, usually around week two of the dissertation module, where the excitement of "I want to write about something important" collides head-on with a blank proposal form asking for a title. That collision is where most people lose their footing. Not because they lack ideas, but because they've never been taught the difference between an interest and a question.

I once sat with a student we'll call her Priya, because that was roughly her situation who was certain she wanted to write about AI in newsrooms. She could talk for twenty minutes about chatbots, hallucinated quotes, algorithmic bylines. Fascinating stuff. But when her supervisor asked what she was actually going to investigate, she froze. She had a subject the size of an entire industry and no way of holding it in two hands. That's not a lack of intelligence. It's a very ordinary, very fixable gap.

Why Interest Isn't the Same as a Topic

An interest tells people what excites you. A topic tells people what you're going to examine, using what evidence, over what period, and among whom. Those are different jobs, and conflating them is the single most common reason proposals bounce back from supervisors with red ink all over them.

Here's the part nobody says out loud enough: sounding clever isn't the same as being specific. "AI in journalism" sounds academic because it genuinely is an academic subject. It just isn't a project-sized one. The subject is the ocean. Your job is to find the one current worth tracing through it.

Testing Whether an Idea Can Actually Be Studied

A useful trick, one I'd genuinely recommend writing on a sticky note above your desk, is to ask what a sceptical examiner would say back to you. If your idea is "editors are quietly using AI without telling readers," the obvious follow-up is: telling readers where, measured how, across which outlets? If you can answer that in a sentence, you've got something real. If you can't, you've got a feeling dressed up as a plan.

This is also where most people misjudge what "difficult" actually means. A crowded subject like misinformation or Gen Z news avoidance isn't hard because it's unoriginal plenty of brilliant work gets done in crowded fields. It's hard because vague framing gets punished faster there. Everyone's already said the obvious thing, so your version has to actually say something.

Students searching for genuinely workable journalism research topics tend to assume the hard part is finding an unclaimed corner of the subject nobody else has touched. It rarely is. The harder, more useful skill is learning to narrow whatever you're already drawn to until it becomes something you could realistically finish, defend, and be proud of by the deadline.

Making a Broad Subject Small Enough to Hold

Narrowing isn't the same as shrinking something until it fits a word count, which is a mistake I see constantly. Good narrowing keeps whatever made you curious in the first place and simply strips away everything that was scaffolding around it.

Take cross-border disinformation as an example. Instead of asking the enormous question of how disinformation spreads globally, ask how one platform's specific moderation policy shaped the spread of one particular claim across two named countries in a defined window. The curiosity survives entirely intact. What's changed is that you can now actually go and study it, rather than gesturing vaguely at a phenomenon too large for anyone to properly examine in a single project.

A habit that genuinely helps: write your working idea as a sentence containing a verb compares, tracks, evaluates, explains instead of a noun phrase sitting there doing nothing. "News avoidance among young adults" describes a subject. "Explains why local newsroom closures correlate with falling attendance at civic meetings in one region" describes a piece of work. Only one of those can be marked.

Reading Before You Decide, Not After

Most students treat background reading as something that happens once the topic's locked in, which is backwards. Reading is what helps you lock it in properly. Skim a dozen decent papers or well-reported pieces in your rough area before committing to anything, and one of two things will usually happen: you'll find your exact angle's already been done thoroughly, or you'll spot a specific gap that hasn't.

This step also tells you something practically important whether the material you'd need even exists. Plenty of promising ideas about platform politics or algorithmic fact-checking quietly collapse because moderation logs, internal policy documents, or reliable audience figures simply aren't accessible to an undergraduate. Far better to learn that during a week of reading than during the panicked third week of a project going nowhere fast.

Knowing When to Let an Idea Go

An idea usually needs changing when you can't explain, in a single sentence, what your work adds that wasn't already obvious to anyone paying attention. Another reliable sign is a methodology that keeps shifting under you one week it's interviews, the next it's content analysis. That's rarely a methods problem. It's almost always a sign the underlying question was never quite settled.

None of this should feel like failure, because it isn't one. Nearly every strong dissertation went through at least one version that got cut down or pointed somewhere else entirely. The students who finish with genuine confidence aren't the ones who guessed correctly first time. They're the ones willing to put an idea under real pressure before they've spent months defending it, rather than after.

Where This Actually Leaves You

By proposal time, you should be able to answer three things without hesitating: what exactly you're studying, what evidence you'll use to study it, and why the answer matters to someone beyond the person marking it. If any answer feels foggy, the idea isn't ready yet no matter how genuinely interesting the subject behind it is.

Priya, in the end, didn't write about AI in newsrooms generally. She wrote about how three regional UK outlets disclosed (or quietly didn't disclose) AI-assisted content to readers over a six-month stretch. Same curiosity that started the whole thing. Just finally small enough to actually hold.