How to write a doctoral dissertation with InkPhD
A month-by-month working method for a medical PhD: from the topic and the reading plan to the chapter tree, the sub-studies, the statistics, the defence rehearsal and the printed volume. Written for candidates who have a supervisor, a deadline and a committee.
Your first week in InkPhD
Do these six things before you write a single paragraph of Chapter 1. Each links to the detailed section below.
- 01
Set up the thesis project
Type "Doctoral dissertation", topic, specialty, language.
- 02
Attach the guidelines
Composite Dissertation checklist + the guideline of each sub-study.
- 03
Fix the research question
PICO, hypothesis, aims and the defensible claim of novelty.
- 04
Build the core corpus
40–80 full texts: landmark papers, guidelines, your own sub-studies.
- 05
Build the chapter tree
Structure Builder or import a published thesis outline.
- 06
Write the literature review first
Tiered: field → controversy → your niche.
Stage 1 — Set up the thesis project
A dissertation is a single InkPhD project that lives for years. Set it up once, properly: the type decides that you get the chapter workflow, sub-sections, versions and the composite checklist.
- 01New project → type "Doctoral dissertation" (this and "Book" are the only types with the full chapter/sub-section workflow).
- 02Title: the working thesis title from your individual plan. Description: 3–6 sentences — population, design of each sub-study, main outcome, what the thesis is supposed to prove.
- 03Writing language: pick the language the volume is written in. The interface stays English; the generated text follows this setting and can be changed per generation.
- 04In Settings fill the front matter now: university, faculty, department, degree, supervisor(s), year, city, abbreviations list. These feed the title page, declaration and lists in every export.
- 05In Settings → Research memory let InkPhD learn your style (citation density, abbreviations, recurring concepts) — it stays as a soft preference and never overrides evidence.
Supervisor noteDo not open a second project for "the article from chapter 3". Sub-studies belong in the same project so the evidence library, numbering and readiness scores stay consistent.
Stage 2 — Two levels of reporting guidelines
A thesis is reported on two levels: the volume as a whole, and each sub-study by its own design. InkPhD ships a composite 42-item Dissertation checklist exactly for the first level.
- 01Set "Dissertation (composite)" as the primary guideline of the project — it covers the structured abstract, native-language summary, list of original publications, tiered literature review, ethics, statistics, participant flow, baseline tables, cross-sub-study coherence, strengths/limitations, future prospects and Vancouver references.
- 02For every empirical chapter run the EQUATOR check with the matching guideline: CONSORT for the RCT, STROBE for the cohort/case-control, PRISMA for the systematic review, TRIPOD+AI for the prediction model, STARD for diagnostic accuracy, CARE for case series, ARRIVE for animal work.
- 03Run the compliance check per chapter, not once at the end. Missing items can be auto-drafted from your library and then edited.
- 04Keep the checklist output — many faculties and journals ask for the completed checklist as an appendix.
Supervisor noteIf a chapter fits no guideline, that is a warning sign about the design — not a reason to skip the checklist.
Stage 3 — Research question, hypothesis, claim of novelty
Committees fail theses on a vague question far more often than on bad prose. Write the question, the hypothesis and the claim of novelty before the reading plan.
- 01In the "Research Question" tab write the main question in PICO form plus 3–5 specific aims, each mapped to one sub-study.
- 02State the hypothesis as a falsifiable sentence, with the direction and the effect size you consider clinically relevant.
- 03Write one sentence of novelty: "No previous study has …". Then test it — run the Novelty Checker later on the literature review to see whether your own corpus contradicts that claim.
- 04Generate the structured abstract only at the very end; generate the native-language summary from it (required by most European faculties).
Supervisor notePaste the exact wording from your approved individual plan into the description — the whole thesis should stay inside that scope.
Stage 4 — Build the thesis corpus (40–80 full texts)
The corpus is the thesis. A dissertation library is bigger than an article library and needs to be organised from day one.
- 01Start with the non-negotiables: current international guidelines, the 5–10 landmark trials, the most recent systematic reviews and meta-analyses, your department's own publications.
- 02PubMed tab: run the advanced search per aim (MeSH + study type + date), import the hits, then use "Fetch full texts" in the Library tab to pull every legally open-access full text.
- 03Upload the PDFs you have legal access to but which are not open access — full text, not the abstract.
- 04Zotero users: connect a read-only key in Settings and sync the collection you already built for the thesis.
- 05Mark 2–3 excellent theses or papers as "style reference" — their voice guides tone, never facts.
- 06If the thesis includes a systematic review, do it in the PRISMA screening module: import the search, screen with the Smart Queue (active learning), record exclusion reasons, and push the included records into the library with one click. The PRISMA flow diagram and the Methods paragraph come out of the same module.
Supervisor noteAdd sources per chapter as you go and attach them to that chapter. A 300-source undifferentiated library produces bland retrieval for every chapter.
Stage 5 — The chapter tree
A medical dissertation usually runs 6–9 chapters with two levels of sub-sections. Build the tree before writing; changing it later costs weeks.
- 01Writing tab → Structure Builder → "AI Generate": it proposes 6–12 chapters with descriptions, rationale, subtopics and PubMed-ready MeSH keywords, based on your title, specialty, question and hypothesis.
- 02Or "Import from Reference PDF": upload a published dissertation from your faculty and InkPhD rebuilds its outline up to three levels — the fastest way to match local expectations.
- 03Typical skeleton: Introduction · Literature review · Aims and objectives · Materials and methods · Results (one sub-chapter per sub-study) · Discussion · Conclusions · Contributions · References · Appendices.
- 04Use sub-sections (demote/promote buttons or drag onto the middle of a card) — the export builds the table of contents from this hierarchy.
- 05Normalise the order once at the start; every new chapter then gets a clean number.
Supervisor noteShow the generated tree to your supervisor before writing. One 20-minute conversation here saves a month of rewriting.
Stage 6 — The literature review, written in tiers
The review is where committees decide whether you know your field. Write it in three tiers instead of one flat summary.
- 01Tier 1 — the field: epidemiology, current standard of care, the guidelines. Tier 2 — the controversy: where the evidence disagrees, and why (design, population, endpoints). Tier 3 — your niche: what is missing, i.e. the gap your thesis fills.
- 02Create one sub-section per tier, attach the relevant sources to each, and generate with 3 000–4 000 words and Hybrid retrieval.
- 03The prompt is tuned to argue, not to list: every sub-section names the study design, reports effect sizes with 95% CI, weighs disconfirming evidence and ends in a synthesis.
- 04Then run Research Gap Detector on the review — it tells you what your own corpus covers but your text ignored, with a draft paragraph and citations.
- 05Close with an explicit paragraph "what remains unknown" that hands over to your aims.
Supervisor noteNever let the review be a chain of "Author X found …". Each paragraph must end with your own interpretation.
Stage 7 — Methods and statistics per sub-study
Methods must be reproducible by a stranger. Write one Methods sub-section per sub-study and keep the statistics explicit.
- 01Per sub-study state: design, setting and period, eligibility criteria, sampling and sample-size calculation, variables and their definitions, measurement instruments, data handling, statistical plan (tests, software, significance level, handling of missing data, correction for multiple testing), ethics approval number and consent procedure.
- 02Use the Statistical Reporting Checker on every Results chapter: it audits 10 items (sample size, 95% CI vs p-only, multiple testing, missing data, subgroups, adverse events, baseline balance, primary vs secondary outcomes, exact p-values, ITT vs per-protocol) and quotes your own text back at you.
- 03Numbers, tables and figures belong in the project assets so they are numbered consistently and exported into DOCX/PDF with captions and the lists of figures and tables.
- 04If the thesis includes software you wrote, use the "Python project" tab: connect the repository, extract verifiable facts with file-and-line provenance, and draft the system description — no invented metrics, and a proper Software reference block in the bibliography.
Supervisor noteWrite Methods while you are still collecting data. Reconstructing it a year later is how protocol deviations get forgotten.
Stage 8 — The daily writing loop
One sub-section per working session. The loop is always the same, which is what makes a thesis finishable.
- 011) Attach sources to the sub-section. 2) Generate (Hybrid, 2 000–3 500 words). 3) Read against the citations. 4) Rewrite the interpretation in your own voice with Smart Continue and the selection toolbar. 5) Re-verify citations. 6) Approve the section as human-reviewed.
- 02Use Fast mode when you only want to see whether the structure and the sources hold; regenerate properly afterwards.
- 03Ask the Copilot the questions your committee will ask: "which of my sources contradict this?", "is this sample size defensible?", "what is the weakest claim in this chapter?"
- 04Version history is per section — snapshot before any large rewrite, and compare afterwards.
- 05Keep the AI Coach visible: it ranks which chapter to work on next by words, evidence, citation trust, grounding and compliance.
Supervisor noteApprove sections as you finish them. An unapproved chapter is an unread chapter, and it will show at the defence.
Stage 9 — Citations, grounding and originality
For a thesis, citation integrity is not a nicety — it is the thing a hostile reviewer will attack first.
- 01Citations render as numbered Vancouver [1], [2] with a full reference list per section; the Citation Verifier gives a trust percentage and marks anything it cannot resolve as "needs review".
- 02Run Claim Grounding per chapter: every atomic claim is labelled supported / partial / unsupported with the supporting excerpt. Fix the unsupported ones by adding the right source or deleting the claim.
- 03Run the Originality check per chapter (semantic similarity against your own library) — it catches paragraphs that stayed too close to the source while paraphrasing.
- 04Before submission export the bibliography in Vancouver (or your faculty's style) and check that the number of references in the list matches the highest number used in the text.
- 05Keep the AI-assistance disclosure in the export if your university requires it — InkPhD stores per-section metadata (AI-generated, human-reviewed, evidence support) for exactly this statement.
Supervisor noteA single fabricated or mis-attributed reference can sink an otherwise good thesis. Treat every "needs review" flag as a blocker.
Stage 10 — Readiness and defence rehearsal
Do not go to the pre-defence with a draft you have not attacked yourself.
- 01Run the Readiness Score per chapter (7 weighted signals → one 0–100 number with priority-sorted action items), then the Project Readiness Overview for the whole volume.
- 02Use the buckets (ready / almost / needs work / major / not audited) to plan the last weeks realistically.
- 03Run the Defense Panel on the Discussion, the Results and the Literature review: five reviewers — Clinician, Statistician, Methodologist, Journal editor, Committee member — return their main concern, likely criticisms, unsupported claims, missing literature and a suggested defence answer.
- 04Rehearse out loud with those suggested answers; then apply the inline suggestions (accept/reject, each anchored to a real quote) to fix the text itself.
- 05Write the Conclusions and the Contributions sections last, straight from what survived all of the above.
Supervisor noteAbove a readiness score of ~85 the remaining points usually need new data, not new sentences. Know when to stop writing.
Stage 11 — The volume, the abstract booklet and the publications
The last two weeks are formatting and paperwork. InkPhD produces the submission package, you produce the signatures.
- 01In Exports enable the structural parts your faculty requires: title page, table of contents, list of abbreviations, declaration of originality, list of figures, list of tables.
- 02Export DOCX for the supervisor and the faculty (real tables, embedded figures), PDF for the print shop, LaTeX if your department uses it, Markdown as your own archive.
- 03The abstract booklet (autoreferat): build it from the structured abstract, the aims, the methods summary, the key results, the conclusions, the contributions and the list of publications — then export separately.
- 04Composite / publication-based theses: keep the list of original publications up to date in the front matter, and map each sub-study to its published paper and its own reporting guideline.
- 05Archive a snapshot at every milestone — InkPhD keeps versions, but your own DOCX/PDF copies are what survive any platform.
Supervisor noteAsk your faculty for the formatting regulation before the final export. Margins, line spacing and the declaration wording are local rules no tool can guess.
A realistic 12-month writing plan
Assuming the data are collected, this is a workable schedule for the writing year. Use the Plan tab to turn it into tasks.
- 01Months 1–2: project setup, guidelines, question and hypothesis, core corpus of 40–80 full texts, chapter tree approved by the supervisor.
- 02Months 3–5: literature review in three tiers, one sub-section per week, gap detection and novelty check at the end of each tier.
- 03Months 5–7: Methods per sub-study, statistical reporting check, figures and tables into the project assets.
- 04Months 7–9: Results per sub-study; every chapter gets grounding + originality + readiness before it is considered done.
- 05Months 9–10: Discussion, Conclusions, Contributions; Defense Panel on the three key chapters and inline fixes.
- 06Months 11–12: structured abstract and native-language summary, abstract booklet, full readiness overview, formatting, export, supervisor and pre-defence rounds.
Supervisor noteBook a weekly 90-minute slot with the plan open. Theses are finished by cadence, not by inspiration.
A dissertation is judged on defensibility, not on volume. InkPhD gives you evidence, structure and rehearsal — the scientific responsibility for every sentence stays yours.