Transcribing interviews for your dissertation.

Say you've done eight interviews of forty-five minutes each: around six hours of audio on your phone. Your supervisor says 'write up the interviews and put them in the appendix'. After twenty minutes of typing you have two minutes of conversation on paper, and you start working out how many evenings this is going to take. More than you have left. Before you book those evenings in, choose your form first.

In a verbatim transcript you write down everything exactly as it sounded: filler words, repetitions, false starts, every 'um', sometimes even pauses and intonation. You use it when the exact way something is said matters, for instance in conversation or discourse analysis. The second form drops the fillers and false starts and keeps the same content in tidy sentences: this is a clean read, internationally also called intelligent verbatim. For most dissertations, a clean read is all you need: it reads well and you lose no content. A summarised version lets go of the word-for-word entirely and is closer to minute-taking; you would rarely use it for qualitative research, because you lose your quotes and your nuance.

Whichever form you go for, choose it up front and keep it consistent across all your interviews. And transcribing is only one step; the full route, from recording to analysis, is below.

As soon as you process interviews with identifiable people, the GDPR rules apply. A recording with a name in it is personal data, and that brings three obligations your supervisor and your institution's data protection officer can check.

Consent: informed consent

Ask for consent beforehand and set down what you are recording, what you will use it for, how long you will keep it and that the respondent can withdraw. Universities want that consent to be 'unambiguous, freely given and specific', documented in writing or verbally (source: Ghent University). The general rules are on can you record a meeting.

Anonymous versus pseudonymous

Replacing names with 'Respondent 1' feels anonymous, but as long as you can still work out who the real person is, through a key file or the stored audio, it is pseudonymisation. The Autoriteit Persoonsgegevens, the Dutch data protection authority, treats pseudonymisation as a security measure: the data remains personal data and the GDPR still applies. Only once the link to the person is irreversibly broken is the data truly anonymous and outside the GDPR. For most dissertations that means you hold pseudonymous data throughout your research, and you handle it as carefully as you would the names themselves.

The data processing agreement

This only applies if you use an external transcription service or app for recordings that contain personal data: the GDPR then requires a data processing agreement (GDPR Article 28). You put it in place before you upload the first recording. As the researcher you remain responsible; the service is the processor (source: the Autoriteit Persoonsgegevens).

Three questions before you upload a recording

• Consent documented?Did you ask the respondent for consent beforehand, and document it in writing or verbally?
• Anonymous or pseudonymous?Is the respondent now untraceable (anonymous), or could you still identify them (pseudonymous)? In the latter case the GDPR still applies.
• Data processing agreement in place?If you use an external service, put the data processing agreement in place before you upload the first recording.

Typing it up yourself gives you full control and you hear every nuance, but for a dissertation with eight to ten interviews it adds up to dozens of hours of pure typing. Automatic transcription saves that time, but it is not flawless: speech recognition stumbles over accents, specialist terms and words that sound alike. Allow one and a half to three and a half hours per interview for checking against the recording (source: Eftekhari, 2024). However you transcribe, checking is part of it.

Timings based on commonly cited figures (Amberscript, StudentJob) and the checking time in Eftekhari (2024). Speed, in other words, is no longer the problem; what remains is the question of where your recording ends up while a service converts it. More on that below.

Transcribing gives you the raw text of your interviews, not your conclusions. A transcript is what was literally said; you still have to draw the codes, the themes and the storyline out of it, and that is precisely the step your supervisor marks.

And sometimes you're after something other than a research transcript. If you just want to know how to turn a recorded conversation into text, without the research context, read conversation transcription. If you don't want a word-for-word transcript but a tidy report with the decisions and the action points, typing out the whole conversation is the long way round: look at the meeting report or writing up meeting minutes. And if you're not sure whether you're allowed to record this interview, start at can you record a meeting.

Sources.

• Scribbr, step-by-step plan for conducting and processing interviews (in Dutch): the three transcription forms and the order of the steps.
• Scribbr, coding qualitative interviews (in Dutch): open, axial and selective coding.
• Scribbr, an introduction to thematic analysis (in Dutch): grouping codes into themes (Braun and Clarke).
• Autoriteit Persoonsgegevens, the Dutch data protection authority (in Dutch): pseudonymised data remains personal data; the data processing agreement (GDPR Article 28).
• Ghent University, informed consent and the GDPR (in Dutch): the consent requirements for research.
• Eftekhari, H. (2024), Transcribing in the digital age: European Journal of Cardiovascular Nursing 23(5); the duty to check and common error patterns.
• Amberscript and StudentJob (in Dutch): typical timings for manual transcription.

General information about research methods and the GDPR, not a legal or methodological verdict. Follow your own programme's requirements and, for sensitive data, the guidance of your institution's data protection officer or ethics committee. Sources consulted in June 2026.