In a qualitative study there are no test statistics to vouch for the results, so reviewers and examiners lean on the methods section. They read it to find out whether they can follow what you did, whether the choices fit your question, and whether the findings could plausibly have come out of that process. This guide covers a structure that works for interview studies, what readers check first, the mistakes that cost the most credibility, and how to describe AI-assisted coding without overclaiming.
A structure that follows the research process
Follow your university's or journal's template where one exists; it takes priority. Where it leaves room, an order that mirrors the project lets readers find every step:
- Design and rationale. Why interviews, and which kind (semi-structured, narrative, expert), given your research question.
- Participants and sampling. Who was invited, by what criteria, how you reached them, how many took part, and who is missing.
- Data collection. Interview guide, setting, length, recording, and any changes after a pilot interview.
- Transcription and preparation. The transcription conventions, how transcripts were checked, and how identifying details were replaced. If you need the detail, see how to transcribe qualitative interviews.
- Analysis. The named method, the steps you took, how the codebook or themes were developed, and how they changed.
- Rigour and reflexivity. The quality criteria you apply and the concrete measures behind them, plus your own position in relation to the topic.
- Ethics and data handling. Consent, storage, pseudonymisation, deletion, approval where required.
What reviewers check first
Most readers run through the same questions. Does the method match the question? Is the sample justified, and is its size argued rather than just stated? Could someone follow the analysis step by step? And does the results section show what the methods section promised? A methods section that describes a three-level coding scheme next to results that only report a handful of themes will be noticed.
Reporting guidelines can serve as a checklist. COREQ is a 32-item checklist for interview and focus group studies, and SRQR is a 21-item standard for qualitative research more broadly. Journals in health and social science often ask for one of them, and even when yours does not, going through the items shows you what readers may expect to find.
Typical mistake: describing the textbook, not your study
The most common weakness is a methods section that paraphrases the method's original description. It tells the reader how thematic analysis or content analysis is supposed to work, and says little about what you actually did. Qualitative analysis is iterative, so your real process probably moved back and forth between steps. Say so, and say why. Another frequent gap is the missing reason: "semi-structured interviews were conducted" is a description, while "because the question concerns how people interpret a process, and the interviews needed room for follow-up" is a justification.
A third problem is silence about limits. A recording that failed, a sample drawn from one organisation, or a participant who dropped out belong in the text. Stated limits read as care; discovered ones read as concealment.
Describing the analysis so it can be followed
Name the method and cite its source, then describe your own use of it. Say whether codes came from theory, from the data, or both. State how many passes you made over the material and what changed between them, and how the codebook was documented. If more than one person coded, report how disagreements were handled; if you calculated agreement, say which measure you used and how the segments were defined. Our article on intercoder reliability with AI coding explains the options and their limits. For the coding process itself, see how to code qualitative data.
Do not claim a criterion you did not meet. If you say the analysis was validated by member checking, there should be a description of who checked what. If you did not do it, name the quality measures you did use instead.
What software can and cannot do here
You write the methods section yourself, because it documents your decisions. Software can make some of the decisions easier to document and to check. In Themera every code is backed by a verbatim quote, and quotes that are not exactly in the material are rejected automatically, so each coded claim can be traced to the text. A coverage indicator shows which responses received no code, the codebook can be edited and the material re-coded, and an intercoder agreement check (blind re-coding, Cohen's kappa) lets you compare your own coding with the AI's. Names can be pseudonymised before analysis, including inflected German names, and results export to DOCX and CSV. Data is stored and processed in the EU (Mistral AI, Paris), not used for training, and a data processing agreement is available.
What software cannot do is decide whether your sample was appropriate, whether a theme is meaningful, or whether your interpretation holds. Nor can it tell you what your institution requires for ethics review or data protection; ask your supervisor or the data protection office.
Disclosing AI-assisted coding
If AI took part in coding, say so in the analysis section, not in a footnote. State what the tool did, what you did, and how you checked its output: for example that you reviewed each code against its supporting quote, edited the codebook, and compared a subset against your own independent coding. Avoid wording that makes the tool the analyst. The interpretation is yours.
Wording to adapt
The text below is a template with placeholders, not a description of a real study. Delete every sentence that does not match what you did.
We conducted [number] semi-structured interviews with [participant group], selected by [sampling approach] to [reason]. Interviews lasted [range] minutes, were audio-recorded with written consent and transcribed in [clean/full] verbatim; automatic transcripts were checked against the recordings. Names and identifying details were pseudonymised before analysis.
Data were analysed using [method, citation]. The codebook was developed [deductively / inductively / both] and revised in [number] rounds; changes were recorded in a codebook log. [If applicable: Initial coding was carried out with an AI-assisted tool. Each code was linked to a verbatim excerpt, and we reviewed every code against its excerpt, revised the codebook, and independently coded [share] of the material to compare agreement ([measure, result]).] Quality was addressed through [specific measures]. The study was [approved by / exempt from review at institution]; data are stored [location] and will be deleted [timing].
Conclusion
A good methods section tells the story of what you did in the order you did it, gives a reason for each major choice, and is open about deviations and limits. Write notes on your decisions while you work, and the section mostly assembles itself at the end.
See a sample analysis with codes, verbatim evidence and the coverage indicator, or try it with your own material: three analyses free, no card needed.