Using AI Interviews in a Dissertation: A 2026 Guide for Master's and PhD Students

Written by Fengming Liu (UCL) · Reviewed by Prof. Shubin Yu (HEC Paris) · October 2026 · Updated: 2026-10-12 · 9 min read

Short answer: you can use AI-moderated interviews in a dissertation when your supervisor agrees, your ethics board approves the AI step, and you report it as part of your method. An AI interviewer is a data-collection instrument: it asks your guide's questions and follow-ups, records the answers as transcripts, and leaves the analysis and the interpretation to you.

When AI interviews suit a dissertation

Semi-structured interviews with a fixed guide, where every participant should get the same core questions.
Participants spread across time zones or countries, or who can only take part in the evening.
Studies in more than one language. QualiTaTi runs interviews in its 10 interface languages.
Larger samples than you could schedule and transcribe alone in a semester.

When they do not

Ethnography, life-history and other designs that depend on a long relationship with participants.
Participants without reliable internet access or who are not comfortable with a chat or voice interface.
Vulnerable groups, unless your ethics board has approved the AI step explicitly.
Interviews where you need to read body language or adapt in ways a written guide cannot anticipate.

Evidence is still building. In a randomised comparison with university students, Wuttke et al. (2024) found AI-led interviews produced data of roughly the same quality as human-led ones on the same questionnaire. They also listed areas where AI interviewers still need to improve. Treat that as a reason to pilot, not as proof for your topic.

Before you start: three conversations

Your supervisor. Agree that AI interviews fit your research question, and how you will report them. You can invite your supervisor to the interview project so they can read the guide and the transcripts.
Your ethics board. Describe the AI step in your application: what the AI does, where the data is stored, and how participants consent. The ethics approval kit has an AI-use disclosure, an information-sheet section, consent statements and a data-management paragraph you can adapt.
Your institution's AI policy. Many universities now publish rules on AI in assessed work. Check whether yours treats AI-assisted data collection or coding differently from writing.

Step by step

Write your interview guide, or paste your research questions and let QualiTaTi draft one for you to edit.
Choose how participants answer: by typing, by voice, or both (see voice interviews for qualitative research).
Run the interview yourself, then with two or three classmates. Revise any question the AI misreads or follows up badly.
Share the link with participants once the guide and the ethics paperwork are final.
Transcripts appear in the project as interviews finish. Read the first few early and adjust if needed.

What a dissertation study costs

A standard interview uses 1 credit a minute, whether participants type or speak. The realtime voice conversation uses 5 credits a minute. Accounts that use the optional EU AI region pay a premium on standard interviews.

Study	Credits	Plan that covers it
Pilot: 3 text interviews of 10 minutes	30	Free plan (30 credits a month)
15 text interviews of 30 minutes	450	One month of Student (€9, 600 credits)
15 standard voice interviews of 30 minutes	450	One month of Student
15 realtime voice interviews of 30 minutes	2,250	About two months of Researchers (€19, 1,200 credits a month)

The Student plan asks for a university email address, or a student ID if you register with another address. Prices as published on the pricing page on 2026-10-12.

Analysing the transcripts

In QualiTaTi: import the interviews into a QDA project and code by hand, or accept and edit AI code suggestions. Keep a record of which codes you changed; examiners will ask.
In NVivo, ATLAS.ti or MAXQDA: if your university licenses one, export a .qdpx file and continue there. See the guides for NVivo, ATLAS.ti and MAXQDA.
Reliability: invite a second coder to code a sample independently; the workspace reports Cohen's kappa (Cohen, 1960).
Thematic analysis: whatever the tool, the analytic decisions are yours. Braun and Clarke (2006) remains the standard reference for the process.

Writing it up

The QDA workspace drafts a methods paragraph from what it recorded: the number of interviews, the AI models and settings used, how codes were produced, and agreement statistics. It also gives a software citation. Edit it into your own words and add what the software cannot know: sampling, recruitment, ethics approval and your analytic approach.

Report the limits too. In text mode the AI sees no tone of voice or body language. In voice mode answers are transcribed automatically, and transcription errors are a known weakness of AI voice interviewing (Tirumala et al., 2025). Say how you checked transcripts.

Checklist for your methods chapter

Why AI-moderated interviews suit your research question.
The guide, the probing setting and how you piloted it.
Ethics approval reference, consent procedure and data storage.
Interview mode (text or voice), number of interviews and their length.
How you coded, who checked the coding, and the agreement figure if you used a second coder.
The software and version, cited.

Frequently asked questions

Will examiners accept AI-moderated interviews?

That depends on your field and institution, which is why the supervisor and ethics conversations come first. Examiners mainly look for a justified choice, transparent reporting and a careful analysis.

Can my supervisor see the interviews?

Yes. Invite them to the interview project by email as a collaborator.

Can I combine AI interviews with interviews I ran myself?

Yes. Upload your own transcripts to the same QDA project and code everything together. Report the two modes separately in your method.

Where is the data stored?

In the European Union. The trust page lists the providers that process data and the data processing agreement your ethics board may ask for.

What if a participant prefers a human interviewer?

Offer an alternative, such as a video call with you, and transcribe it into the same project. Note the mix in your method.

Sources, checked 2026-10-12: Wuttke et al. (2024), AI Conversational Interviewing: Transforming Surveys with LLMs as Adaptive Interviewers, arXiv:2410.01824; Tirumala, Jain, Leybzon and Buskirk (2025), Mic Drop or Data Flop? Evaluating the Fitness for Purpose of AI Voice Interviewers, arXiv:2509.01814; Braun and Clarke (2006), Using thematic analysis in psychology, Qualitative Research in Psychology 3(2); Cohen (1960), A coefficient of agreement for nominal scales, Educational and Psychological Measurement 20(1).