The thesis workspace that cannot invent a source.
Your document, your papers, and a copilot that can only cite what is actually in them.
3.2 Data collection and analysis
For the analysis of the interview material a qualitative approach was chosen, because the research question asks how participants describe their own decision-making.
Thematic analysis was selected as the method. It offers a procedure that is not tied to a pre-existing theoretical frameworkBraun & Clarke, 2006, p. 78, which means the six phases can be applied across different epistemological positions.
Coding proceeded inductively in the first pass and was checked against the interview guide in a second, deductive passMayring, 2022, p. 97. Both passes were carried out by the author; a second coder was not available, which is addressed as a limitation in section 5.4.
Saturation was reached after the fourteenth interview, which is consistent with the range usually reported. Recruitment stopped at that point.
Transcripts were prepared using simple orthographic transcription, since the analysis is concerned with the content of what was said rather than the fine-grained features of its deliveryFlick, 2020, p. 412.
This chapter reads as human-written.
Checked 2,214 of 2,340 words. About 3 in 100 human-written academic texts get flagged anyway — a clean result is reassurance, not proof.
Two passages sit too close to their original.
Both are cited, so this is not plagiarism. It is the kind of paraphrase a supervisor recognises, and it is quicker to fix now than to defend later.
Opens your actual document
Your .docx or LaTeX project — formatting, styles and your supervisor's comments intact.
Your library, not the open web
Drop in PDFs, or import a Citavi or Zotero library. It reads them instead of recalling them.
Citations that resolve
Every claim links to the passage it came from. Open the page, read the line, keep writing.
Both checks, always on
AI and overlap, in words rather than a percentage, while you write instead of the night before.
Saved as you write
Your document and your sources live in the cloud, versioned — not on one laptop that can be lost, stolen or die in submission week.
Upload as many papers as you have
No cap on documents or storage. A thesis reading list runs to hundreds of PDFs, and a tool that rations them is no use.
It says no.
Ask any other assistant for a citation and you get one: right shape, plausible journal, a DOI that goes nowhere. This one searches your library — and when your library does not support the sentence, it says so, then offers the version you can defend instead.
A citation you cannot open is worse than none. You are the one in the viva.
“Saturation was reached after 14 interviews” — find me a source.
I will not invent a citation for “fourteen”.
The paper students usually cite here is not in your library, so I cannot show you the line. Two of your papers touch saturation. Neither gives you fourteen.
Flick (2020) · p. 155
Saturation is not a threshold that can be fixed in advance; it is a decision made in the field, and the researcher is obliged to make the grounds for it explicit.
Suggested instead: say the decision was made in the field — that you can support. You can report fourteen without citing it.
The assistant wrote most of this. Rewrite the section in your own words — half an hour now, a different conversation in the viva.
You stay the author.
Every paragraph carries how much of it is yours. When the assistant has written most of a section, the workspace says so and asks you to put it back in your own words — while that still takes half an hour.
This is the part other writing tools have no reason to build. They are paid to write more of your thesis. We would rather you hand in something you can defend, because the same check that reads your chapter is the one your examiner will run.
The file you hand in is clean.
Every source resolves
Nothing is cited that the assistant could not open in a paper you uploaded. No invented DOIs, no journal that does not exist.
No invisible characters
Text pasted in from AI tools carries zero-width characters and homoglyphs. They do not belong in a submitted file, so the workspace strips them out.
Your AI declaration, written as you go
58 of 64 German universities require you to declare AI use, and several want the prompts. The log builds itself instead of being reconstructed the night before.
Removing invisible characters is formatting hygiene — it makes a clean file, and we do not claim it changes what any detector concludes about your writing. What changes that is the balance above.
Get early access
An address, nothing else. No name, no university, no supervisor — the same as everything else we do.