After three or four years, you know every sentence of your thesis and can no longer read any of it fresh. Your supervisor has seen earlier versions, your co-authors know their own article, and the assessment committee will be the first to read the whole thing as strangers. This page sets out what a critical reader looks for in a doctoral thesis, part by part, and where theses most often slip.
No one sees their own blind spots, with or without AI help. Esmond points them out in your draft from within your field, along with factual errors and inconsistencies.
A PhD thesis is assembled over years. The introduction was drafted early and revised late, the middle chapters were each finished in their own season, and the conclusion was often put together in the last weeks before submission. Each part made sense when it was finished. Whether the parts still say the same thing is a separate question, and it is the one you are least able to answer after months inside the text.
Most feedback you get along the way is partial by design. A supervisor reads the chapter in front of them, a journal reviewer reads one article, a co-author reads the analysis they worked on. The reader who takes in the abstract, the introduction and the conclusion in one sitting arrives at the very end, and in many countries that reader sits on the committee.
A hypothetical example of a commentChapter 3 reports an association between workload and sick leave in one hospital. Your conclusion says workload causes sick leave in hospitals. Which of the two does your design support?
A first comment on one part of your own draft, right away and without an account.
Get a free previewReaders of a PhD thesis tend to read like peer reviewers for a journal: they ask whether the contribution is visible and whether the evidence carries it. The table lists the questions that come up in almost every field, with a check you can run on your own draft.
| What is read for | The question behind it | Check in your draft |
|---|---|---|
| Contribution | After this part, what does the reader know that was not known or not shown before? | Can you point to the sentence that says it? |
| Positioning | Does the literature show where the debate is open, or does it list one source after another? | Does each paragraph of the review end in a position, or in a summary? |
| Claim discipline | Does the evidence in this part carry exactly the claims that rest on it? | Take your three strongest claims and note the evidence for each next to it. |
| Method | Why can this approach answer the question, and why not an alternative? | Is the choice argued, or only described? |
| Limitations | Are they used to set the reach of the contribution? | Does each limitation return where you interpret the findings? |
| Consistency | Do definitions, assumptions and terms hold across the part? | List your key terms with their first definition, then check every later use. |
The problems that matter most in a PhD thesis often sit between its parts, which is why a chapter can look sound on its own and still raise questions in the committee. Four patterns come back across fields.
A monograph carries one argument from the first page to the last, so the risk is a chapter that drifts away from it. A thesis by publication brings together articles that were each built to stand alone, so the risk is that nobody, including the candidate, says how they fit. Guidelines at the University of Oslo, NTNU and the University of Helsinki ask for a separate summarizing part that brings the articles together. The details, including co-authored articles and the summary that frames them, are on the page about a thesis by publication.
A reader who holds a history thesis to the standard of a clinical trial is reading it wrong, and so is the reverse. The questions in the table above stay the same; what a good answer looks like changes with the field.
| Field | What a strong part shows | A question a reader in the field may ask |
|---|---|---|
| Medicine and biomedical sciences | Reproducible methods and claims that stay within the design | Does the conclusion keep the difference between association and effect? |
| Psychology | Measures that fit the construct and an honest reading of effect sizes | Is a null result reported as what it is, without being read as proof of no effect? |
| Social sciences | Transparent analysis and, in qualitative work, reflexivity | How far does a finding from these cases or this sample reach? |
| Law | Normative and conceptual precision | Is the standard used to judge the law stated and defended? |
| Engineering and computer science | Validation that holds outside the test setting | Does the evaluation support the general claim about the system? |
| Humanities | Interpretation and command of sources | Is the reading argued against the obvious alternative reading? |
The criteria your committee applies are rarely secret. The University of Helsinki tells its doctoral candidates: “Getting acquainted with the instructions aimed at the examiners is a recommendable idea in general, as it will give you a good idea of how your final doctoral thesis will be evaluated.” Check whether your university publishes guidance for its assessors, find it early, and read your draft against it, part by part.
A practical order: read the abstract, the introduction and the conclusion in one sitting, before anything else. Note every claim in the conclusion and look for the place in the chapters that carries it. Then read each chapter on its own for the questions in the table. What is left after both passes is what to raise with your supervisor. For the last weeks, see before your PhD defense.
This example is made up. A candidate in public health has three empirical chapters: a survey in one region, interviews with nurses in two hospitals, and a register study. Each chapter is careful about its own limits. The conclusion, drafted in the final month, states that workload drives early retirement among nurses and recommends national staffing ratios. Read on its own, every chapter holds. Read next to the conclusion, two questions arise: where does the thesis show a causal link, and where does it move from two hospitals and one region to the national level? The candidate keeps the recommendation, ties it to the register findings, and states what the interview chapter can and cannot add to it.
Esmond evaluates every assignment before you submit. It never writes it.
Esmond is an AI tool that doesn't write for you. If your course doesn't allow any AI tools, don't use Esmond for that work. If you're unsure, ask your instructor or advisor.
Esmond reads one part of your thesis per upload, up to about 125 pages: a chapter, an article, the introduction or the general discussion. At upload, choose the dissertation option; it is meant for doctoral work. Esmond reads the part from within your field and flags factual errors, inconsistencies and blind spots, as annotations in your own PDF and a report with questions a committee might ask, usually within a few minutes. Within the part you upload, it points out where a claim goes further than the evidence in that part supports. Upload the closing chapter and it is read as a part in its own right: whether its claims are carried by what it reports, and whether its terms keep their meaning.
Esmond adds no text and gives no verdict on the thesis as a whole: whether the work is substantial enough for a doctorate cannot be judged from one part, and that judgment belongs to your supervisor and your committee. See how Esmond works and the sample report.
Only if it fits within about 125 pages. Most doctoral theses are longer, so upload one chapter, article or part per review. Each review covers only the part you upload.
No. It gives no verdict on the outcome and no rating. It points out where a part raises questions, so you can work on them before your supervisor and committee read it.
The article itself usually stays as published. Comments on it are most useful for what you add around it: the summarizing part, the introduction and the general discussion, where you can still explain and update.
That depends on your university and sometimes on your supervisor. Check the rules for your doctoral program, and ask your supervisor if you are unsure.
Read on November 7, 2026.
Get your paper evaluated before you submit