Resources

Start here: what do you need to do?

Most of what you need is already on this site, and it is organized by problem rather than by topic. Below is the fastest route to the right page, and the outside resources we actually use.

For supervisors

Questions Supervisors Ask — diagnostic answers to the questions Chairs and committee members actually ask, with a curated reading list for doctoral supervisors. Free.

Finding and checking literature

Google Scholar — free, and the starting point for most searches. Set up your library links so it shows what your institution can get you.

Scopus and Web of Science — subscription databases, usually through your library. Also the practical test for whether a journal is legitimate: if a journal is indexed in neither, ask why.

Predatory Journals — how to tell, and why it matters more than people think.

The adjacency trap — how many sources you actually need, and verification. See also The Literature Review.

Writing and style

Scholarly Style — what scholarly writing actually is, and the habits that get flagged.

Clarity — the foundation, and our one-word summary of the whole business.

Purdue OWL — free, comprehensive, and the best general reference for APA and other styles. We use it ourselves.

Your institution's own style requirements — which override everything above. Get them in writing.

Statistics and analysis

G*Power — free power analysis software, and the standard. Run it before you write your proposal, not after it is approved. See Methodology at Proposal Stage.

SPSS, NVivo, and equivalents — usually available free through your institution. Ask before buying anything.

AI tools

We have written more about this than about anything else, because it changed everything and the guidance elsewhere is thin.

Human-in-the-Loop Structural Refinement — what these tools are genuinely for, and where they fail. Navigating AI Detection Paranoia — what a detector score measures, and what to do if you are flagged. Purging Citation Contamination — fabricated references, and how to find them. Academic Honesty — where the line is.

When something has gone wrong

Stalled Dissertations — nothing is moving and you do not know why. Rejected and Bounced Dissertations — it came back. Panic — out of time and cannot think. Real or Scam? — how to tell whether a service is legitimate before you pay anybody.

Our own writing

Blog — longer pieces on doctoral research, writing, and the thinking underneath both.

Publications — our founder's own publication record, in full.

Writing worth reading as writing

Read scholarly articles for their prose, not only for their findings. These three are here because of how they are written.

Labib, S.M.; Lindley, S.; Huck, J.J. Estimating multiple greenspace exposure types and their associations with neighborhood premature mortality: A socioecological study. Science of the Total Environment, 789, 147919.

Classic scholarly register, third person throughout, and heavily cited. Read the methods section and notice how much gets specified without the sentences becoming long. This is what a committee means when it asks for scholarly writing.

Helbich, M.; Burov, A.; Dimitrova, D.; Markevych, I.; Nieuwenhuijsen, M.J.; Dzhambov, A.M. Sociodemographic inequalities in residential nighttime light pollution in urban Bulgaria: An environmental justice analysis. Environmental Research, 262, 119803. Open access. Read it

First person plural — we analyzed, we speculate, we found. Same department at the same university as the paper above, written to the opposite convention.

Read them one after the other, and then notice the rule underneath. Helbich uses the first person where a human made a decision — we used stratified sampling, we reclassified age, we fitted two models. The passive appears where a procedure ran — associations were assessed, the analyses were conducted in R.

That is the useful distinction, and it is not the one candidates are usually given. Not never use I. Rather: name yourself where you made a judgment somebody could have made differently, and stand back where the method simply ran.

Opportunities and Barriers for Renewable Energy in NAFTA. Howse, R. and van Bork, P. Renewable Energy and International Law Project, CEC Research Paper, February 2006.

A different kind of writing, and it is still held to the same standard.

Written for policymakers rather than for a journal, and almost everything about the surface changes. There is no methods section. The audience is named and served — a boxed explanation of what intermittent electricity is and why it matters, because a policymaker will not know and needs to. There are exclamation marks: at least when the sun shines or the wind blows!

What does not change is the rigor.

Every factual claim is footnoted — a hundred and eighteen of them, to government data, industry associations, and primary sources. Opinion is marked as opinion: it is the authors' opinion that, the authors are skeptical of, we do not believe that. The reader is always told which is which.

And notice the seam. Parts I and II are industry and policy overview in plain prose. Part III is legal analysis, dense with cases run inline — EC-Asbestos, US-Shrimp, Methanex. You can see the register change at the boundary, in one document, by the same authors.

If you are writing a capstone or a professional doctorate, this is closer to your register than either paper above. And it is the answer to a candidate who thinks that writing clearly means writing unrigorously.

It is the other way round. A dense sentence can hide a claim you have not supported. A clear one cannot. Rigor lives in the sourcing and in marking your judgments as judgments — and clarity is what lets a reader check both.

One more thing worth seeing in the Helbich paper, and it is not the prose. Look at the section headed Declaration of generative AI in scientific writing. That is a required heading in a major Elsevier journal, and it reads, in full: No generative AI and AI-assisted technologies were used during the preparation of this work.

Look also at the CRediT authorship statement, which itemizes what each author actually did. Writing – original draft. Methodology. Formal analysis. Data curation.

Disclosure is not a hypothetical. It is a heading, and somebody signs it.

What AI actually did to real text

Printable: the nine failure modes, with tick boxes — valid as of August 2026. The cases are below.

We keep a log. Every case here happened during ordinary work — on this site, on client material, or on published sources — with the date, the model named, and the passage before and after.

When AI edits a methods section — two precise sentences from a published paper read as a duplication. That paper had six authors and went through peer review. Every reader left both sentences standing, because every one understood why both were needed.

When an AI forgets what it told you — a real degree reported as an invented term, by the model that had researched and added it weeks earlier. It described its own verified work as a hallucination.

When it misreports what just happened — a summary that reversed who had done what, written from a record still on screen.

These are current, paid, frontier-tier models. Not free tools and not older versions.

What they have in common: the output reads perfectly well. Nothing announces the loss. Every case was caught by a person who had the previous version or who remembered what had been there — which is the argument for keeping dated drafts, and it is the whole of the defense.