How to write a PhD proposal

Over the past weeks, I have been reading a number of proposals for potential candidates to our PhD program. Reading proposals is always interesting: you get a glimpse of what prospective doctoral candidates are excited about, which problems they see in the field, and how they imagine spending the next years of their lives.

But this year, I noticed something that worries me.

In the age of AI, many proposals have become louder. They are full of buzzwords. They contain long lists of activities. They combine fashionable methods, trendy terminology, and ambitious claims in one sentence after another. Yet, somehow, the scientific core is often missing.

A PhD proposal is not a product pitch. It is not a grant application written by throwing every current keyword into the introduction. It is not a promise to test everything, model everything, optimize everything, and then somehow make sense of the results.

A strong PhD proposal starts somewhere else: with a real research question.

By a real research question, I mean a question that can make a dent in the universe. Perhaps the dent is small. Most PhD contributions are small. But it should still be a scientific contribution: original in scope, technically meaningful, and transferable beyond one narrow case.

Here is what I am looking for when I read a PhD proposal:

• Start with the literature, not with the tools. Before you decide to use AI, machine learning, numerical modeling, experimental work, or any other method, sit with the literature. Read deeply. Think carefully. Ask yourself: what do we not understand yet?
• Formulate one clear main research question. The proposal should not be a collection of activities. It should revolve around one central, fundamental question that guides every choice you make and that will move your field forward.
• Move beyond “let us try this and report what happens.” For example, I am not very interested in yet another concrete mix where another constituent is added and the results are reported. I am interested in understanding why the material behaves the way it behaves, so that the insight can be translated to other mixes and used to predict the effect of other constituents in the future.
• Aim for transferability. Your contribution should not only work for one dataset, one material, one structure, or one case study. The best proposals ask a question that can help us understand a broader principle. Returning to concrete mixes: I want to understand why a change in the mix or curing methods will change the outcomes, so that I can predict how other mixes will behave too.
• Keep the objectives focused. A proposal needs one clear objective and, depending on the duration of the PhD, two or three sub-objectives. More is not better. More often means that the project has not yet been thought through.
• Make the schedule realistic. In a three-and-a-half-year PhD, the experimental work cannot take three years. In our context, I would generally not expect more than one year of lab work. The remaining time is needed to analyze, model, write, publish, and bring the project together.
• Let the research question determine the methods. Do not propose to test every possible property and then try to make sense of the results afterwards. Turn it around. If you expect that a certain parameter controls the behavior of the material, explain why that parameter matters – and limit the experimental work to testing what you need to understand.
• Make the parts coherent. Experimental work, analytical modeling, numerical simulations, and data analysis should not be separate boxes in the proposal. They should speak to each other. Each method should help answer the main research question.

A good PhD proposal is not flashy. It is thoughtful. It shows that the candidate has spent time with the literature, has identified a real gap, and has the discipline to turn that gap into a manageable research project.

So, if you are writing a PhD proposal, resist the temptation to impress with buzzwords. Impress with clarity. Show that you understand the mechanics of the problem. Show that you know what needs to be measured, modeled, or analyzed, and why.

A PhD is not about doing everything. It is about making one original contribution that others can build on.

And in times when AI can generate pages of polished text in seconds, careful thinking has become even more valuable. I am not looking for a proposal that sounds sophisticated. I am looking for a proposal that shows scientific judgment.

What, in your experience, makes a PhD proposal truly convincing?