InkPhD
DISSERTATIONS · RESEARCH ARTICLES · SYSTEMATIC REVIEWS · GRANT PROPOSALS

From first draft to publication-ready.

InkPhD is the scientific writing OS for medical research — dissertations, research articles, systematic reviews, clinical protocols, grant proposals, and more. Evidence-grounded text with strict [1] [2] citations, EQUATOR compliance, and a Pre-publish Readiness Score that tells you whether your manuscript would survive peer review or a defense — before you submit.

Start writing your manuscriptSee the workspace
11
EQUATOR guidelines
0
Hallucinated cites
5
AI defense reviewers
inkphd.com / workspace / phd-pancreatitis
Literature Review · 6,210 words

2.3 Surgical management of severe acute pancreatitis

Open necrosectomy was the historical standard, with mortality reported between 23 % and 38 % across high-volume centers[1][2]. The minimally invasive step-up approach reduced new-onset multi-organ failure by an absolute 21 % (95 % CI 8–34, p = 0.002)[3].

Subsequent meta-analyses, however, found no statistically significant difference in 90-day mortality[4], suggesting that the benefit derives from avoided complications rather than survival per se.

[3] van Santvoort HC, et al. Step-up vs open necrosectomy. N Engl J Med. 2010.
[4] Bakker OJ, et al. Endoscopic vs surgical drainage. JAMA. 2012.
BUILT FOR THE STANDARDS OF
Harvard MedicalKarolinskaUniversité ParisCochraneNEJM AIEQUATOR Network
EVIDENCE ENGINE

The same prompt, two universes apart.

Plain LLMs invent confident citations to studies that don't exist. InkPhD only cites papers in your own evidence library, with verifiable DOIs and PMIDs.

Generic LLM · No grounding Risk

"Open necrosectomy carries a mortality rate of approximately 36 %(Smith et al., 2019), whereas the minimally invasive approach reduces mortality to 12 % (Johnson 2021). A recent meta-analysis confirms this finding(Chen et al., 2023; n = 4,210)."

  • Smith et al. 2019 — paper does not exist
  • Chen 2023 (n = 4,210) — fabricated sample size
  • 95 % CI and p-value omitted
InkPhD · Evidence-grounded Verified

"Open necrosectomy was historically associated with mortality between 23 % and 38 % across high-volume centers[1]. The step-up minimally invasive approach reduced new-onset multi-organ failure by 21 % (95 % CI 8–34, p = 0.002)[2], though 90-day mortality did not differ significantly[3]."

  • Every [n] resolves to a real DOI in your library
  • Effect sizes copied verbatim with 95 % CI
  • Cross-checked against Crossref + NCBI

InkPhD pre-loads your dissertation's evidence pack into per-chapter ChromaDB collections, then forces the language model to cite ONLY from that pack — using the exact Manubot tokens provided by the retrieval layer.

THE WORKSPACE

Document-centered. Three panes. Zero context-switching.

Hierarchy, editor, intelligence — laid out as a research instrument, not a chat window.

LEFT PANE

Section Hierarchy

Drag-and-drop reorder up to 3 levels of subsections. Each section owns its own evidence library — strict retrieval isolation prevents leakage across topics.

  • Sub-section nesting up to depth 2
  • Library badge per section
  • AI Structure Builder (auto-proposes structure in 8s)
CENTER PANE

Scientific Editor

Vancouver-style numbered citations are rendered live. Tab to accept ghost-text continuations. Continue button finishes the current paragraph without breaking structure.

  • Vancouver / APA 7 / AMA 11 / BibTeX export
  • Live ghost-text Jenni-style editor
  • Selection toolbar: expand · shorten · academic tone
RIGHT PANE

Intelligence Panel

Scientific Copilot, Inline Evidence Inspector, Citation Verifier, EQUATOR snapshot — five live tabs running in parallel against the section you are writing.

  • AI Copilot · grounded in your library
  • Verifier · Crossref + NCBI cross-check
  • Reviewer · severity-scored issues with auto-fix
ONE WORKSPACE · TEN PROJECT TYPES

Whatever you're writing, the engines adapt.

Pick your project type at creation and the entire workspace reconfigures — guideline catalogue, structure templates, prompt persona, and readiness criteria all shift accordingly.

PhD

Doctoral Dissertation

Multi-chapter thesis with composite EQUATOR guideline · sub-section nesting · defense simulation.

RA

Research Article

IMRAD structure for original research · target a specific journal · abstract + figures.

SR

Systematic Review

PRISMA-compliant · auto-extracted PICO tables · risk-of-bias scaffolding.

MA

Meta-analysis

Forest-plot ready · effect-size aggregation · heterogeneity reporting.

CP

Clinical Protocol

SPIRIT-compliant trial protocols · ethics statement · CONSORT-ready outline.

GR

Grant Proposal

Specific aims · innovation · approach · with budget-justification scaffolding.

CR

Case Report

CARE-compliant single-patient narrative · timeline · informed-consent reminder.

CA

Conference Abstract

Structured 250–300 word abstract for ESMO / ASCO / NEJM AI etc.

BK

Book / Monograph

Chapter-based long-form · same hierarchy + section libraries as dissertation.

ED

Editorial / Commentary

Short opinion piece grounded in your evidence library, with hedged scholarly voice.

STATISTICS & DATA

Upload your dataset. Get the analysis, the figure and the sentence that reports it.

No exporting to SPSS and pasting screenshots back. Your CSV or Excel file becomes real survival curves, model tables and the wording a reviewer expects — computed from your data, inside the manuscript.

KAPLAN-MEIER · LOG-RANK

Survival analysis

Kaplan-Meier curves with the classic number-at-risk table beneath the plot, median survival, log-rank comparison and survival at your chosen horizons.

COX HR · SCHOENFELD

Cox models you can defend

Hazard ratios with 95% CI plus an automatic proportional-hazards check (Schoenfeld residuals, per-covariate and global p-values) and a diagnostic log-log plot.

ROC · AUC · YOUDEN

Diagnostic accuracy

ROC curves, AUC with 95% CI, Youden cut-off with sensitivity and specificity — publication-ready in colour or print black & white.

β · OR · r

Regressions & correlations

Linear β and logistic odds ratios with confidence intervals, baseline and group-comparison tables, Pearson/Spearman matrices with significance markers.

AUTO CROSS-REFERENCES

Numbering that never breaks

Every figure and table carries a stable marker. Reorder or delete one and every in-text mention renumbers itself — in the editor, in Word (SEQ/TOC fields) and in the PDF.

METHODS · AUTO-DRAFT

The Methods paragraph, written

InkPhD drafts the "Statistical analysis" paragraph from the analyses you actually ran and drops it straight into your Methods section.

PREMIUM ENGINES

Six independent analyzers. One readiness score.

Every analyzer runs in parallel against your manuscript and reports back into a single weighted Pre-publish Readiness audit.

THE MOAT · PRE-PUBLISH READINESS

Will my manuscript survive peer review?

A single 0-100 audit aggregating Citation Verifier · EQUATOR · Statistical Reporting Checker · Research Gap Detector · Novelty · Library Coverage · Length. Action items are deduped and priority-sorted.

92/100 Citations78/100 EQUATOR64/100 Stats
runs in 30-60 s
4 LLMs in parallel
persisted per chapter
COMPLIANCE

11 EQUATOR guidelines · 426 checklist items

CONSORT, STROBE, PRISMA, TRIPOD+AI, STARD, CARE, SPIRIT, CHEERS, ARRIVE, CONSORT-AI, CLAIM — plus a custom 42-point DISSERTATION composite curated from ICMJE + Vancouver.

DEFENSE PREP

5 AI reviewers · simulated panel

Clinician · Statistician · Methodology · Journal · Committee — each returns headline concerns, unsupported claims, and suggested defense responses.

PAPERQA2-STYLE

Claim Grounding · per-claim verdicts

Each atomic claim retrieved, judged supported / partial / unsupported. Unsupported claims get a one-click 'start refinement' suggestion.

WAVE D

Stats · Gaps · Novelty

10-item statistical reporting audit, prioritized research-gap detection, and novelty scoring 0-10 per substantive claim — with verbatim quotes.

InkPhD Deep Writer

What five generations actually buy you.

Everything below is the unedited output of one real run on our demo project — a Discussion written from 8 indexed papers: 3,294 words, 4 parts, 20 paragraphs audited, 86% citation trust. Three passes instead of one prompt is why a Deep Writer run counts as five generations.

Researcher: 5 / month · Doctoral & Lab: unlimited

Real output · Pass 2 · Draft · Step-up vs open necrosectomy in necrotising pancreatitis

The randomised evidence most directly addressing the first issue is PANTER, which compared a percutaneous-drainage-first step-up pathway with primary open necrosectomy and demonstrated a substantially lower risk of the composite endpoint of major complications or death with step-up (40% vs 69%; risk ratio 0.57; 95% CI 0.38 to 0.87; p = 0.006), while mortality was similar (19% vs 16%) and the between-group difference was largely explained by fewer new-onset multiple-organ failure events (12% vs 40%; p = 0.002) in the step-up arm. [1]

TENSION, a multicentre RCT comparing endoscopic step-up against surgical step-up, found no superiority of the endoscopic route on the primary composite of death or major complications (43% vs 45%; risk ratio 0.97; 95% CI 0.62 to 1.51; p = 0.88), with identical mortality (18% in both groups), but fewer pancreatic fistulas (5% vs 32%; p = 0.0011) and shorter hospital stay (mean 53 vs 69 days; p = 0.014) with endoscopic management. [2]

[1] van Santvoort HC, et al. A step-up approach or open necrosectomy for necrotizing pancreatitis. N Engl J Med. 2010. doi:10.1056/NEJMoa0908821

[2] van Brunschot S, et al. Endoscopic or surgical step-up approach for infected necrotising pancreatitis. Lancet. 2018. doi:10.1016/S0140-6736(17)32404-2

Verbatim excerpts from a real run, shortened only for length. Your output depends entirely on the papers in your own library — the engine cannot cite anything you have not indexed, and any claim it cannot ground is flagged instead of written. Method note: the outline stage follows the STORM method (Stanford) and the verification stage the OpenScholar method (AI2); both are implemented inside InkPhD, which is not affiliated with either project.

See which plan includes it
DEFENSE READINESS

The five questions your committee will ask. Before they ask them.

Run a parallel simulation against five distinct reviewer personas — each trained to surface the gap your draft is hoping you won't notice.

CLINICAL.RELEVANCE

The Clinician

Senior attending

"What is the absolute risk reduction in your high-grade subgroup?"

STATS.METHODOLOGY

The Statistician

Biostatistics chair

"Multiple testing correction is not stated; α inflated to ~0.17."

STUDY.DESIGN

The Methodologist

Cochrane editor

"Selection bias is plausible; sensitivity analysis recommended."

EDITORIAL.RIGOR

The Journal

NEJM associate

"Abstract overstates novelty relative to van Santvoort 2010."

DEFENSE.READINESS

The Committee

Defense chair

"Where in your thesis do you address the contradicting Bakker 2012 result?"

WHY INKPHD

The part other AI writing tools leave to you.

Reference managers find papers. Editors polish language. InkPhD is built for what happens between the data and the defence.

Statistics inside the manuscript

Survival, Cox with assumption testing, ROC/AUC and regressions computed from your own dataset — not a chat about statistics.

Defence simulation

Five expert reviewer personas interrogate your draft and hand you the answers to rehearse before the committee does.

Pre-publish Readiness Score

One honest number from seven signals: citation trust, EQUATOR, statistical reporting, evidence gaps, novelty, density and length.

EQUATOR compliance with auto-fix

11 guidelines and 426 checklist items — including a DISSERTATION composite — with retrieval-grounded fixes for what is missing.

Every citation verified

Each DOI and PMID is checked against Crossref and PubMed. Unresolved references are flagged, never quietly printed.

Word output a university accepts

Real Word tables, native SEQ/TOC fields, lists of figures and tables with page numbers, title page and declaration — plus LaTeX.

Code-to-Methods provenance

Attach a Python project and get Methods text with file-and-line evidence for every technical claim.

PRISMA screening end to end

Dedupe, AI-assisted title/abstract decisions with reasons you can overrule, the flow diagram and the study-selection paragraph.

Why not just use a general chatbot

InkPhD (retrieval-grounded) vs ChatGPT / Gemini (general chat)

General assistants are excellent at language. They are not built to prove that a sentence is supported by a specific paper — which is the one thing a reviewer, a supervisor and an editor will test.

 InkPhD · RAG platformChatGPT / Gemini · general chat
Where the facts come fromOnly from papers you uploaded or imported — retrieved, ranked and quotedThe model's training data and whatever the user pasted
CitationsNumbered Vancouver [1], [2] resolved to DOI/PMID and re-verified against your libraryOften plausible-looking but unverifiable; fabricated references are a known failure mode
What happens when evidence is missingThe claim is flagged "unsupported" and you are told which search would fill the gapThe sentence is written anyway, in the same confident tone
Reporting standards11 EQUATOR guidelines + a dissertation checklist, audited item by itemNo notion of CONSORT, STROBE, PRISMA or SPIRIT compliance
Statistics in your textAudited: 95% CI vs p-only, multiple testing, missing data, ITT vs per-protocolNo audit — numbers are reproduced as written
Structure of a long documentSection tree with per-section libraries, versions and a readiness score per sectionOne long chat thread; structure and consistency are the user's problem
Review before submissionDefence Panel: clinician, statistician, methodologist, editor, committee memberGeneric "improve my paper" feedback
Deep multi-pass writingInkPhD Deep Writer: multi-perspective outline → per-subsection draft → claim-by-claim verification (Researcher plan and above)One prompt, one answer — no critique or re-retrieval loop
Your manuscriptStored in your project, never used for training, exportable and erasable on demandDepends on the provider and the plan

InkPhD uses the same frontier models underneath — the difference is what they are allowed to say, and what has to be proven before it reaches your manuscript.

HONEST COMPARISON

InkPhD next to the AI writers and the discovery engines.

Seven tools that get mentioned in the same breath but solve different problems. Prices as published in June 2026. Where they beat us, we say so.

1 · Against the AI writing tools

1 · Against the AI writing tools
CapabilityInkPhDJenniPaperpalStudyTexter
Price€0–279 / month$12–30 / month$12–29 / month€69–249 once per paper
What it is really forWriting and defending a medical or scientific manuscript on your own evidenceFast AI drafting with autocompleteLanguage editing and submission polish for journalsGetting a finished thesis draft written for you
Disciplines coveredMedicine and the wider sciences — clinical medicine, biology, pharmacy, public health, dentistry, nursing, engineering and social sciences: EQUATOR (CONSORT, STROBE, PRISMA, SPIRIT, TRIPOD…) for clinical work, IMRAD / PRISMA / grant structures for everything elsegeneral academic writing, no methodology layergeneral academic English, any fieldgeneral academic writing
Retrieval-grounded drafting from YOUR librarypartial — cites from its own libraryno — edits the text you already havepartial — writes from sources it picks
Numbered citations verified against Crossref / PubMedpartial — citations inserted, not DOI-verifiedpartial — checks style, not existencepartial — 2 853 styles, no DOI verification
Unsupported claims underlined in the editoryes — claim confidence——
Statistics from your own datasetyes — Kaplan-Meier, Cox + PH check, ROC/AUC, regressions———
Reporting-guideline (EQUATOR) complianceyes — 11 guidelines, 426 items, auto-fix—partial — journal submission checks—
Defence / reviewer simulationyes — 5 reviewer personas + readiness score—partial — pre-submission review report—
Figure & table numbering with native Word fields——partial — formatted document
Word / Google Docs add-inno — plannedpartial — Word add-inyes — Word, Docs, Overleaf, browser—
ExportDOCX · PDF · LaTeX · Markdown · HTMLDOCX · LaTeX · BibTeXDOCX · LaTeXDOCX / PDF of the delivered thesis

2 · Against the literature discovery engines

2 · Against the literature discovery engines
CapabilityInkPhDElicitConsensusSciSpaceResearchRabbit
Price€0–279 / monthFree tier · $10–169 / monthFree tier · $9–12 / monthFree tier · $12–20 / monthFree
Disciplines coveredMedical + scientific: clinical medicine, biology, pharmacy, public health, engineering, social sciences — with the reporting guideline that fits each designall fieldsall fieldsall fieldsall fields
Index you can searchPubMed + OpenAlex + Semantic Scholar + Crossref + arXiv125M+ papers — the strongest discovery engine here200M+ papers via Semantic Scholar280M+ papersSemantic Scholar graph
Consensus meter (how many studies agree)yes — design-weighted, with the quote it judgedpartial — claim summariesyes — that is its whole point——
Citation network mapyes — internal citations + co-cited hubs + suggested reads——partial — connected papersyes — that is its whole point
PRISMA screening + flow diagramyes — dedupe, AI screening, diagram, paragraphyes — built for large-scale screening—partial—
Custom data-extraction columnspartial — fixed PICO findings tableyes — any columns you define, exportable—partial—
Read-along PDF reader with notesno — planned——yes — chat and highlight inside the PDF—
Writes a full, citation-verified manuscriptno — extractions and summariesno — answers questionspartial — short drafts—
Your own dataset → figures & statistics————
Open-access full text fetched automatically by DOIyes — Europe PMC, Unpaywall, OpenAlex, arXivpartial——

Where Elicit is better

Its index of 125M+ papers finds work you would otherwise miss, it screens thousands of records at a time and you can define your own extraction columns with CSV export. InkPhD retrieves from the sources you add, and our screening is built for tens to hundreds of records. If discovery is your bottleneck, run Elicit alongside InkPhD.

Where Jenni and Paperpal are better

Jenni is faster and lighter for pure drafting, and Paperpal lives inside Word, Google Docs and Overleaf — the distribution we do not have yet — with strong language editing for non-native speakers. InkPhD asks more of you up front, because the output has to survive a defence.

Where SciSpace and ResearchRabbit are better

SciSpace has a genuine read-along PDF reader with highlights and per-paragraph chat; ResearchRabbit maps the literature graph better than anyone and is free. Our map covers your own library plus its co-cited classics, not the whole graph.

Where StudyTexter is better

You pay once and receive a finished draft in days, with a human editor available. It asks very little of you. InkPhD asks for your data, your sources and your judgement on every claim — deliberately slower, because the result has to be genuinely yours.

Where InkPhD is better

Nothing else here computes survival curves, Cox models with a proportional-hazards check or ROC curves from your dataset, verifies every numbered citation against Crossref and PubMed, runs EQUATOR compliance with auto-fix, turns a draft audit into a tracked writing plan, renumbers figures with native Word fields and then interrogates your manuscript like a review panel.

What we do not do

No exam revision aids, no ghost-writing service, no plagiarism check against a global corpus (ours is semantic, against your own library), and no Word add-in yet. We would rather say it here than let you discover it after paying.

Compiled June 2026 from each vendor's public pricing and feature pages. Features change — tell us if anything here is out of date and we will correct it.

Write in your language

Generation in 12 languages, evidence in any language

Choose the output language per project or per section. Your sources can be in any language — retrieval and citations work regardless, and the interface stays English so co-authors and supervisors always see the same thing.

  • enEnglish
  • bgBulgarian
  • deGerman
  • frFrench
  • esSpanish
  • itItalian
  • ptPortuguese
  • plPolish
  • roRomanian
  • elGreek
  • trTurkish
  • ruRussian
HOW IT WORKS

Four steps. From empty workspace to defense-ready manuscript.

01

Build the evidence pack

Upload PDFs or pull papers from PubMed with MeSH-filtered search. Each document is chunked, embedded with OpenAI text-embedding-3-small, and assigned to sections.

02

Structure the manuscript

AI Structure Builder proposes sections with rationale, sub-topics, and MeSH-ready PubMed search terms — adapted to your project type (dissertation, research article, systematic review, grant proposal, and more).

03

Write with the engines on

Sections generate under your selected EQUATOR guideline with strict Vancouver citations. Continue, ghost-text, selection toolbar — all evidence-grounded.

04

Run the readiness audit

One click runs Citation Verifier + EQUATOR + Stats + Gaps + Novelty + Reviewer simulation. The score and action items tell you exactly what to fix.

PRICING

Honest pricing. Every tier ships with the same anti-hallucination guarantee.

Annual saves you two months. Prices load straight from our billing system, so this page and checkout can never disagree. VAT and invoice details are on the Billing & Invoices page.

Starter

Try the evidence-grounded workflow on one project.

€0forever
Get started free
  • 1 project
  • 10 sources
  • 5 AI generations / month
  • Markdown & HTML export

Student

For one thesis or paper on a student budget.

€12/ month

or €120 / year — save €24 (2 months free)

Start writing
  • 1 project
  • 150 sources
  • 40 generations / month
  • DOCX & PDF export
  • Statistics & figure numbering

Researcher

For a single paper or review, from sources to submission.

€39/ month

or €390 / year — save €78 (2 months free)

Start writing
  • 5 projects
  • 300 sources
  • 150 generations / month
  • 1 Python project
  • DOCX & PDF export
  • Readiness score & claim grounding
  • InkPhD Deep Writer — 5 premium sections / month
RECOMMENDED

Doctoral

Full dissertation workload with defence preparation.

€89/ month

or €890 / year — save €178 (2 months free)

Start writing
  • Unlimited projects
  • 2 000 sources
  • 600 generations / month
  • 5 Python projects
  • Defence simulation
  • Priority generation queue
  • Unlimited InkPhD Deep Writer runs

Research Lab

Shared evidence libraries for a whole group.

€279/ month

or €2790 / year — save €558 (2 months free)

Contact sales
  • Everything in Doctoral
  • 5 seats
  • Unlimited sources & Python projects
  • Shared evidence libraries
  • Onboarding session

Payments handled by Mollie (cards, iDEAL, Bancontact, SEPA, Apple Pay). VAT applied under EU OSS rules; reverse-charge invoices available for businesses with a VAT number. Billing, invoices & right of withdrawal

Reviewed by the people who use it

What authors say about InkPhD

0.0from 0 reviews

The first reviews are being moderated. Used InkPhD? Leave yours.

TRUST & COMPLIANCE

Built for the paperwork your institution will actually ask for.

Ethics committees and research offices ask three questions: who processes the data, is the AI use declared, and can the author prove human oversight. Here are the answers, in writing.

EU AI Act

Not high-risk, not a medical device. Article 50 transparency: AI involvement is always explicit, every AI section is labelled, and each claim is traceable to its source.

Read the statement

GDPR

Self-service data export and permanent erasure in the app. No model training on your manuscripts. Rights answered within 30 days.

Read the statement

Article 28 DPA

A signed data-processing agreement for universities and hospitals, with security measures, breach timelines and audit support.

Read the statement

Subprocessors

Every provider that touches your data is listed with purpose, data minimisation and location. 30 days notice before changes.

Read the statement

Cookies

Session and workspace preferences only. No advertising, no analytics, no cross-site tracking.

Read the statement

Research Use Only

No diagnosis, no treatment recommendations, no clinical decision support. The RUO notice is rendered on every export.

Read the statement

Operator: International Sci Ink Press Ltd · Bulgaria, EU · contact form

SAFEGUARDS

A writing instrument. Not a medical device.

Research Use Only

InkPhD does not provide clinical decision support. RUO disclaimer rendered on every export.

AI Transparency

Every chapter records which sections were AI-generated, evidence support score, and reviewer approvals.

Citation Verifier

Crossref + NCBI cross-check. Unverified citations are flagged in-line, not hidden.

No global model training

Your library, your retrieval. We never train a shared model on user content — only retrieval-conditioned personalisation.

Your manuscript deserves a rigorous writing partner.

Start with the Hobby tier — no credit card. Upgrade when you run your first Pre-publish Readiness audit and see what's missing.

Open the workspaceRead AI transparency report

InkPhD stores only what is strictly necessary to run your session and remember your workspace preferences. No advertising, no analytics, no cross-site tracking. Cookie Policy · Privacy