Nine checks · defense prep · tropa-1 AI score

AI Thesis Checker

Nine dimensions graded out of 100, findings quoted from your own text, a prioritized to-do list, your likely defense questions — and a tropa-1 AI-detection score before you submit.

🔒

Sign in to check your thesis

A thesis review runs on your credit balance, so we need to know who you are before we start. It costs 2 credits per word — a 10,000-word thesis is roughly 20,000 credits.

PDF or DOCX · any languageSee a sample report

See the thesis check in action

A short walkthrough of what you upload, what comes back, and how to read the report.

What the Report Actually Looks Like

Nine graded dimensions, findings quoted from your own text, a to-do list you can work through, and a parsed chapter outline. Everything below is a walk-through of one report.

This is a made-up example. The thesis, the numbers and the quoted passages are invented for this page — not a real student's document, and not a prediction of what your own report will say.

masterarbeit_final_v7.pdf

Adoption of Municipal Open-Data Portals in German Cities · English · Master's thesis

Example output

Overall score

68/100

Weighted roll-up of the nine dimensions below.

Executive summary

A carefully organised empirical study with a strong theoretical chapter and clean referencing. Two things would cost you at a defense: the conclusion states a causal relationship the cross-sectional design cannot carry, and the case-analysis chapter — where your contribution sits — runs on a fraction of the sources the rest of the thesis uses. The sample size is also reported differently in chapters 3 and 5.

18,420

Words

164

In-text citations

Bibliography entries

8.9

Citations / 1,000 words

APA

Citation style detected

Chapters parsed

Nine graded dimensionsEvery dimension scored 0–100 with a written verdict

Nine graded dimensions

score out of 100 · higher is better

Every dimension carries its own score and a written verdict. Three values are labelled here; in the report each one opens into its verdict paragraph and the findings behind it.

Abstract

Solid

Argumentation

Watch

Consistency

Needs work

Chapter Structure

Solid

Logical Errors

Needs work

Source Density

Watch

Citation Style

Solid

Terminology

Watch

Bibliography

Strong

Verdict · Logical Errors

“The empirical work is sound, but the conclusion converts a cross-sectional association into a causal statement, and the sample-size discrepancy between chapters 3 and 5 is never addressed. Both are visible to a first reader.”

FindingsQuoted from your own text, graded by severity, each with a fix

Findings, anchored to your text

20 findings in this example

Each one is graded, quotes the passage it refers to, names the section and page, and ends with a fix you can actually carry out.

Critical2Major5Minor9Suggestion4

CriticalLogical Errors6.1 Answering the research question · p. 71

The conclusion states a causal claim the design cannot carry

“The analysis demonstrates that portal usability drives citizen participation: cities with more usable portals record measurably higher engagement.”

The study compares seven municipalities at a single point in time. That design can establish an association, not a direction — the same data is equally consistent with participatory cities investing in better portals.

Fix

Reword the claim as an association, name the confounders you cannot rule out (municipal IT budget, prior civic-tech activity), and move the causal reading into the outlook as a hypothesis for future work.

MajorSource Density4.3 Cross-case observations · p. 48

Chapter 4 carries your contribution on a tenth of the sources

“The interviews confirm that usability barriers, rather than data availability, explain the low usage figures reported in Section 4.1.”

Chapter 4 runs 4,620 words on 10 in-text citations — 2.2 per 1,000 words, against 17.0 in the theoretical background. Your most contestable claims are the least supported passages in the document.

Fix

Anchor the three cross-case claims in 4.3 to the literature you already reviewed in 2.4, or state explicitly that they are your own findings and not drawn from prior work.

MinorBibliography2.4 Governance models · p. 19

Three works cited in the text are missing from the reference list

“… as shown in earlier work on portal governance (Bauer et al., 2016).”

Three in-text citations have no matching entry in your reference list, and two list entries are never cited anywhere in the text. This is a cross-check of your citations against your own bibliography — not a comparison against a database of published work.

Fix

Add the three missing entries, then either cite or delete the two orphaned ones. Both directions are listed with page numbers in the full report.

Your to-do listPrioritised, with an effort estimate per task

Prioritized to-do list

ordered by impact, tagged by effort

Effort is a rough estimate, not a promise about your evening — but it tells a student with three days left which items are still reachable.

1
Rewrite the causal claims in 6.1 as associations and name the confounders.
Logical ErrorsArgumentationmedium
medium
2
Source the three cross-case claims in Chapter 4, or mark them explicitly as own findings.
Source DensityArgumentationsubstantial
substantial
3
Reconcile the sample size — Section 3.2 says twelve municipalities, Table 5.1 reports fourteen.
Consistencyquick
quick
4
Settle on one term for the object of study and define it once in 2.1.
Terminologymedium
medium
5
Add page numbers to the twelve direct quotations that are missing them (APA).
Citation Stylequick
quick

Chapter outlineLength and citation density, chapter by chapter

Parsed chapter outline

every chapter with its length and its sourcing

Two views of the same outline, side by side: how long each chapter is, and how densely it is sourced. A document-wide average of 8.9 citations per 1,000 words hides what the second column shows.

ChapterLength · wordsSourcing · citations / 1,000 words

1Introduction

1,240

7.3

2Theoretical Background

4,180

17.0

3Methodology

2,050

11.7

4Case AnalysisThin sourcing

4,620

2.2

5Cross-Case Discussion

2,890

13.1

6Conclusion

1,180

4.2

7Limitations & Outlook

780

9.0

04,800018

Bars are scaled to the longest chapter and the densest chapter in this document, so the comparison is within the thesis — not against other people's.

PDF exportThe whole report in a printable layout — no extra credits

Export the whole report as a PDF

Scores, findings with their excerpts, the to-do list, the outline and the defense questions in one file — the format supervisors and writing centres ask for. Exporting costs no credits.

0 credits

Reminder: every figure on this page belongs to an invented sample thesis. We do not publish real customer reports, and none of these numbers describe how any particular document will score.

Will your thesis get flagged as AI?

Universities increasingly run submitted work through AI detectors, and most students only find out what those tools say about their writing after the submission deadline has passed. That is the wrong order.

The uncomfortable part: writing your thesis yourself is not a guarantee of a clean result. Detectors key on formality, even sentence rhythm, clean grammar and scaffolded structure — which is exactly what academic writing is supposed to look like. Careful writers and non-native English speakers get flagged the most, and the manual "tells" people cite make it worse rather than better.

So every thesis check includes a pass from tropa-1, our text-detection model. You get an overall score, a verdict, and the specific sections that scored highest — so if a chapter reads like a machine wrote it, you can look at that chapter while you still have time to do something about it. It is the same engine behind our AI essay detector.

tropa-1 AI score

Example output

41Uncertain

0 · reads human100 · reads AI

Sections that scored highest

4.2 Portal usability across the seven cases

“Furthermore, it is important to note that usability constitutes a central determinant of adoption, as the preceding analysis has demonstrated.”

2.3 Definitions and scope

A high section score is a place to look, not a finding about you. Sample data from the same invented thesis as above.

Where we stop: a detection score is guidance, never proof. It cannot establish who wrote a document, ours included, and we will not tell you a number means you are safe. What it gives you is advance warning and a place to look.

The Nine Checks in an AI Thesis Review

Every dimension is graded 0–100 with a written verdict and concrete findings quoted from your own text — including the page and section they came from.

📝

Abstract

Your abstract is the only part many examiners read twice, and the part most students write last and fastest. We check that it actually states the research question, the method, the result and the contribution — not just the topic — and that the claims in it are ones your chapters really deliver.

🧩

Argumentation

We trace whether the thesis you announce in the introduction is the thesis you defend in the body and the one you conclude with. Common failure: three chapters of solid work that never connect back to the research question, so the conclusion asserts something the evidence never established.

🔁

Consistency

Contradictions between chapters written months apart — a sample size that changes from 42 to 45, a limitation admitted in chapter 5 that chapter 2 denied, a hypothesis quietly reworded. Examiners find these, and they read as carelessness even when the underlying work is sound.

🏗️

Chapter Structure

We map your outline and flag proportions that will draw questions: a 40-page literature review against a 9-page analysis, a methods chapter that appears after the results, missing transitions between chapters, or a section nested three levels deep that holds two paragraphs.

⚠️

Logical Errors

Claims that overreach the evidence, correlation presented as cause, circular reasoning, a generalisation drawn from a sample that cannot support it, or conclusions that quietly assume what they set out to prove. These are the openings an examiner walks through at the defense.

📚

Source Density

Citations per 1,000 words, measured per section rather than for the document as a whole. A healthy average can hide the real problem: a literature review carrying every reference while your analysis chapter runs eight pages with two citations. Thin patches are where supervisors write "evidence?" in the margin.

🔖

Citation Style

We detect which style you are actually using — APA, MLA, Chicago, Harvard, IEEE — and find where you drift out of it: an ampersand here and an "and" there, missing page numbers on direct quotes, inconsistent et al. thresholds. Formatting rarely fails a thesis on its own, but it is the cheapest mark to lose and the first thing a picky second reader counts.

🔤

Terminology

One concept, three names across five chapters — "user", "participant", "subject" — or a key term defined in chapter 2 and used with a different meaning in chapter 4. In a technical thesis this is a genuine comprehension risk, not a style nitpick.

📖

Bibliography

We cross-check your reference list against your in-text citations in both directions: works cited in the text but missing from the list, and entries in the list nothing ever cites. We also flag incomplete entries — missing year, missing publisher, a dead-looking DOI field. This is reference hygiene, not a plagiarism scan against a source database.

One clarification, up front. The source and bibliography checks look at the quality and consistency of your referencing. They do not compare your text against a database of published work, so this is not plagiarism detection and does not replace whatever plagiarism software your university uses.

How the Thesis Check Works

📤

STEP 1

Upload your thesis

Drop in the PDF or DOCX — the whole document, not a summary. Any language; the report comes back in the one you wrote in.

🔍

STEP 2

We parse the structure

Headings, chapters, in-text citations and your bibliography are extracted so every check runs against the real document layout.

📊

STEP 3

Nine graded checks

Each dimension gets a score out of 100 and a written verdict, alongside a tropa-1 AI-detection pass over the text.

✅

STEP 4

Report, to-do list, defense prep

A prioritized to-do list ordered by what costs you most, plus the examiner questions your weak spots invite.

Check your thesis

Walk Into Your Defense Knowing the Hard Questions

Examiners do not ask about your strongest chapter. They ask about the assumption you skipped past, the sample you did not justify, the limitation you buried in a footnote. So we generate defense questions from the weak spots the review actually found — not from a generic list.

Three of the eight questions in the sample report

Example output

“Your methods chapter describes twelve municipalities; Table 5.1 reports fourteen. Which is the study, and does the difference change your result?”

Why they would ask

Section 3.2 and Table 5.1 give different case counts, and the discussion draws on the larger set without explaining the addition.

Targets 3.2 Sample and case selection

“What rules out the reverse reading — that cities with active civic participation invest in better portals?”

Why they would ask

Section 6.1 states that usability drives participation, but the design is cross-sectional and no confounders are discussed anywhere in the thesis.

Targets 6.1 Answering the research question

“Chapter 4 is where your contribution sits, and it cites ten works. Which study in your review would disagree with it?”

Why they would ask

Source density in the analysis chapter is 2.2 per 1,000 words against 17.0 in the literature review — an unguarded chapter reads as an invitation.

Targets 4.3 Cross-case observations

Targeted, not generic

Each question names the section it probes. If your methods chapter never defends the sample size, you get the question an examiner would ask about that sample size — in those words.

With the reasoning attached

Every question comes with the rationale: what in your text invites it. That tells you whether to prepare an answer or go back and fix the underlying gap.

Useful before the deadline

Run the check while you can still revise and half the questions stop being questions. Run it the week of the defense and you at least know where the pressure is coming from.

Included · no extra creditsOptional second pass inside your report

Referenzlage: the literature your sources cite and you do not

A second pass resolves your bibliography against OpenAlex, follows the citations of the works you already cite, and surfaces the ones your own sources repeatedly point at while your thesis never mentions them. The kind of omission a supervisor in the field spots in ten seconds.

You start it yourself from inside the report, and it is gated on honesty: if we cannot resolve enough of your reference list against OpenAlex, we tell you that instead of drawing a picture built on a third of your sources. Coverage is strongest for journal articles with a DOI and weakest for books, German-language and humanities sources — so this will not work equally well for every thesis.

What it looks like in your report

Example output

Bibliography resolved

41 of 63

Below roughly half resolved, no gap list would be shown at all.

Institutional drivers of open government data adoption in local administration

Bauer et al. · 2016

cited by 9 of your resolved sources

A stage model for municipal data platform maturity

Novak & Lindgren · 2019

cited by 7 of your resolved sources

Beyond publication: usage patterns of open data portals

Terzi et al. · 2020

cited by 6 of your resolved sources

Invented placeholder titles. A missing work is a prompt to check, not evidence that anything is wrong with your thesis — and this is still a citation-pattern analysis, not a plagiarism check.

Who the AI Thesis Checker Is For

🎓Bachelor's students

Usually your first document longer than 30 pages. The failure modes are structural — a research question that drifts, a literature review that never feeds the analysis — and they are much cheaper to fix two weeks before the deadline than after the grade.

📗Master's students

More sources, more chapters, more months between the first section and the last — which is exactly how consistency and terminology problems get in. A full pass over the assembled document catches what chapter-by-chapter supervision does not.

🔬PhD candidates

For a dissertation the risk shifts to argumentation across hundreds of pages and to reference hygiene across hundreds of entries. The defense-question list is built for exactly this audience: it probes where your argument is thinnest.

👩‍🏫Supervisors & writing centres

A structured first pass so your feedback time goes to the intellectual content instead of counting citation-style slips. The AI-detection score is a conversation starter with a student, never a verdict about them.

What You Get Back

An overall score and summary

A weighted roll-up of the nine dimensions plus a short executive read on the state of the document — the thing to look at first when you open the report.

Findings quoted from your text

Each issue is graded critical, major, minor or suggestion, and carries the excerpt, the section heading and the page it came from — so you are not hunting for what the report means.

A prioritized to-do list

Ordered by impact and tagged by effort — quick, medium, substantial — so a student with three days left knows what to do with them.

Metrics you can check yourself

Word count, in-text citation count, bibliography entries, citations per 1,000 words, the citation style we detected, and a parsed outline of every chapter with its length and source count.

Defense questions and the AI score

Predicted examiner questions with their rationale, and the tropa-1 detection result with the sections that scored highest — presented as guidance to act on, not a verdict.

A PDF of the whole report

Export everything above as a single PDF to annotate, print or send to your supervisor. The export costs no credits.

Find the problems before your examiner does

Nine graded checks, a to-do list you can work through, the defense questions your thesis invites, and an AI-detection read on your own writing — in minutes.

Check your thesis See pricing

AI Thesis Checker: FAQ

Is this a plagiarism checker?

No, and we want to be precise about the difference. We do not compare your text against a database of published sources, so we cannot tell you whether a passage matches something already in print. What we check is citation and bibliography quality: whether your in-text citations and reference list agree with each other, whether your citation style is applied consistently, whether entries are complete, and whether your sections are supported by enough sources. That is a real and useful check — but it is not plagiarism detection, and you should still run whatever plagiarism software your university requires.

What is the tropa-1 AI score, and is it proof of anything?

It is our text-detection model's estimate of how likely your writing is to read as AI-generated, reported per section as well as overall. It is guidance, not proof. No AI detector — ours or anyone else's — can prove who or what wrote a document, and a high score is not evidence of misconduct. Use it the way you would a spellchecker warning: a prompt to look again at a specific passage before someone else does.

Can my university tell that I used this?

There is nothing in your document for them to find — we analyse the file you upload and return a report to you; we do not edit, rewrite or watermark your thesis, and we do not contact your institution. That said, check your own university's rules on permitted writing support before you use any tool. Policies differ, and that is your call to make, not ours.

Is my thesis stored, and who can see it?

Your upload is processed to produce your report, and the report is kept in your account so you can come back to it. Nobody else gets access to your document, and we do not publish or resell it. You can delete your data, and institutional customers can be set up in zero-retention mode. The full detail is in our privacy policy — read it before uploading unpublished research.

Which languages does the thesis check work in?

Upload your thesis in whatever language you wrote it — the review is not restricted to a fixed list, and the report comes back in the language of your document rather than translated into English. Two honest caveats. First, the structural parsing that finds your chapters and reference list recognises headings in English, German, French, Spanish, Italian, Portuguese and Dutch; a thesis in another language still gets reviewed, and numbered headings are still detected, but an unnumbered heading in, say, Polish may not be picked up, which makes the chapter outline less complete. Second, citation-style detection covers APA, MLA, Chicago, Harvard and IEEE, so it is strongest for the traditions that use them.

Which file types can I upload?

PDF and DOCX. Upload the complete thesis rather than individual chapters — several checks (consistency across chapters, bibliography cross-referencing, chapter proportions) only work on the whole document. A PDF exported from your word processor parses more reliably than a scanned one.

How long does the AI thesis review take?

Minutes, not days. A full-length thesis takes longer than a short paper because every chapter is analysed separately, but you get the report in the same sitting — it is designed to fit inside a revision session, not a supervision cycle.

What does it cost?

The thesis check costs 2 credits per word of the document you upload, so a short paper costs a fraction of a full dissertation. Credits are pay-as-you-go and do not expire; see our pricing for current packs. Exporting your finished report as a PDF is free and costs no further credits.

How is this different from proofreading or a writecheck-style service?

Proofreading services fix language: grammar, phrasing, typography. This is a structural and argumentative review — whether your abstract matches your results, whether your analysis chapter is under-sourced, where an examiner will push at your defense — plus an AI-detection read on the text. They are complementary. Many students want both; this one runs in minutes and returns a scored report rather than a marked-up manuscript.

Will this improve my grade?

We are not going to promise that, and you should be sceptical of any tool that does. Grades depend on your research, your data and your examiners. What the report gives you is a specific, prioritized list of the weaknesses a careful reader would find — fixing them is still your work.

AI detection results are provided as guidance and are not definitive proof of how a document was written. This tool does not check your text against a database of published sources and is not a substitute for your institution's plagiarism software or for your supervisor's judgement. All sample reports, scores and excerpts shown on this page are illustrative examples built from a fictional thesis.