Verbatim or clean verbatim? Transcription rules for theses and research
October 16, 2026 · 6 min read
Interviews in a thesis usually have to follow fixed transcription rules. How clean verbatim and full verbatim differ, what an AI transcript already gives you and what you still add by hand.
If you interview people for a thesis or research project, you need to put them in writing — following rules you describe in your methods section. Most rule sets fall into two families: clean verbatim and full (or true) verbatim.
Clean verbatim
Content comes first. Everything said is transcribed, but smoothed: fillers like “um” and false starts are removed, dialect and contractions are lightly standardized, and stutters disappear. Pauses are marked, speakers get short labels such as I for interviewer and R for respondent, and timestamps appear at the start or end of each paragraph.
Full verbatim
How things are said matters too. Fillers, broken-off words, emphasis, laughter, overlaps and exact pause lengths are recorded. You need this when you analyze conversational behavior — in linguistics, conversation analysis or some psychological research.
What an AI transcript already gives you
| Feature | AI transcript | Your editing |
|---|---|---|
| Literal content | largely there | check names, terms, numbers |
| Smoothing | often happens automatically | usually fits clean verbatim |
| Fillers, stutters | often dropped | add by hand for full verbatim |
| Speaker labels | with a plan: Speaker 1, 2, … | replace with I and R |
| Timestamps | on every passage | convert to the required format |
| Pauses, laughter, emphasis | not included | add by hand |
For clean verbatim, an AI transcript is an excellent basis: most of the work is checking and adjusting, not typing. Full verbatim needs much more manual work, because models drop exactly the features it is about.
A practical workflow
- Transcribe the recording with speaker separation and export to Word.
- Rename “Speaker 1” and “Speaker 2” to I and R with find and replace.
- Compare the full transcript against the recording once; click a timestamp when unsure.
- Apply your department's rules: pause marks, unclear passages, anonymization.
- Anonymize names, places and employers before the transcript goes into your appendix.
State in your methods section that you used an automatic transcript as a basis and checked it fully by hand. That is transparent and matches what many departments now expect.