Skip to main content

Methodological rigour

How to Disclose AI Use in Your Qualitative Methods Chapter (Without Risking Your Viva)

Published 11 June 2026

You used an AI tool somewhere in your qualitative analysis. Maybe it helped surface candidate quotes, or suggested a first pass at codes. Now you're staring at your methods chapter wondering the same thing every postgraduate researcher is asking: how do I disclose this without an examiner deciding my findings aren't really mine?

It's a legitimate fear. Used carelessly, AI in qualitative work is an academic-integrity risk — not because AI is forbidden, but because you can be left unable to show how you reached your interpretations. The good news: disclosure done properly doesn't weaken your thesis. It strengthens it. Here's exactly what examiners and journals expect, and how to write it.

What examiners actually expect (it's not “don't use AI”)

The consensus across journals and examination boards has settled quickly. Using AI in qualitative analysis is increasingly accepted — provided your methods section discloses the tool used and the role it played, and you can demonstrate trustworthiness. A recent framework for trustworthy LLM-assisted qualitative analysis (Jiang et al., “Efficiency with Rigor!”, arXiv:2501.00775) puts it plainly: the AI must function as a transparent, supervised assistant — never an autonomous analyst — with documentation that lets others trace and scrutinise your analytic decisions.

In practice, examiners want four things:

  • Transparency — what tool, what model, what it did.
  • Human primacy — evidence that you made every interpretive decision.
  • An audit trail — a traceable record of how codes and themes were developed.
  • Reflexivity — your awareness of your own role in the interpretation.

These map onto the classic trustworthiness criteria (Lincoln & Guba: credibility, dependability, confirmability) and onto reporting standards your discipline may use — COREQ, SRQR, or, for reflexive thematic analysis, the RTARG guidance (Braun & Clarke).

The principle: assistance is fine; unaccountable assistance is not

The line examiners care about is simple. AI may help you organise, search, and surface — it may not decide what your data means. If a tool auto-generates themes and you “review” them, you're on shaky ground: you can't fully reconstruct why each theme exists. If a tool surfaces candidate evidence and you accept or reject each one — with that decision recorded — you can defend every line of your analysis. Same speed, completely different defensibility.

This is why an audit trail matters more than any single disclosure sentence. A good audit trail records every candidate the AI surfaced, every one you accepted or rejected, when, and why. That record is what turns “I used AI” from a liability into evidence of rigour.

How to write the disclosure — the four parts

A strong AI-disclosure statement in your methods chapter covers:

  • The tool and model — name them, and the version/date if you can.
  • The role — what the AI did (e.g. “surfaced candidate evidence segments for researcher review”).
  • What the AI did NOT do — explicitly: it did not generate themes, make coding decisions, or interpret.
  • The safeguard — every decision was made and confirmed by the researcher, and a complete audit trail of accepted/rejected suggestions is available for inspection.

A worked example you can adapt

AI-assisted evidence retrieval was used during coding. [Tool/model] surfaced candidate text segments semantically related to each code in the codebook; it did not assign codes, generate themes, or interpret the data. The researcher reviewed every surfaced segment and accepted or rejected each one. All interpretive decisions — coding, theme development, and narrative construction — were made by the researcher. A complete audit trail recording each AI suggestion and the researcher's decision is retained and available for examination. No analytic conclusion rests on an unreviewed AI output.

Adapt the bracketed parts to your tool and methodology. For reflexive thematic analysis, add that themes were developed(not “emerged”), that you treated your subjectivity as an analytic resource, and that inter-coder reliability was not sought because coding was reflexive rather than the application of fixed categories. For codebook or content analysis, describe your codebook and any reliability checks instead.

The reflexivity half people forget

Disclosure isn't only about the AI — it's about you. State your position relative to the data and how it shaped your interpretation. AI changes nothing here: the reflexive account is still yours to write, and examiners read it as a marker of methodological maturity.

A faster way to produce all of this

Writing the disclosure by hand is doable — but keeping a complete, examiner-grade audit trail of every coding decision across an entire dataset is the hard part, and it's exactly where most AI-assisted analyses fall down. This is the problem QualIntel OS was built for: the AI surfaces candidate evidence, you confirm every decision, and the platform keeps the audit trail current and generates a non-editable AI-disclosure statement from it — methodology-aware, the way your examiner expects. The analysis, and the credit, stay yours.

Start free

Frequently asked questions

Is it acceptable to use AI in qualitative analysis for a thesis?

Yes — using AI in qualitative analysis is increasingly accepted, provided your methods chapter discloses the tool used and the role it played, and you can demonstrate that you (not the AI) made the interpretive decisions. The risk is not using AI; it is using it in a way you cannot account for.

What should an AI disclosure statement in a methods chapter include?

Four things: the tool and model used; the role the AI played (e.g. surfacing candidate evidence); what the AI explicitly did NOT do (assign codes, generate themes, interpret); and the safeguard — that every decision was made and confirmed by the researcher, with a complete audit trail available for inspection.

Will using AI get my thesis flagged for academic integrity?

Only if you cannot show how you reached your interpretations. If an AI auto-generates themes and you merely review them, that is a risk. If the AI surfaces candidate evidence and you accept or reject each one with that decision recorded in an audit trail, you can defend every line of your analysis.

Do I need an audit trail if I used AI in my analysis?

Effectively yes. An audit trail — a record of every AI suggestion and the researcher's accept/reject decision — is what turns 'I used AI' from a liability into evidence of rigour. It supports the dependability and confirmability criteria examiners assess.

Further reading: Jiang et al., “Efficiency with Rigor! A Trustworthy LLM-powered Workflow for Qualitative Data Analysis” (arXiv:2501.00775); Lincoln & Guba (1985) on trustworthiness; Braun & Clarke on reflexive thematic analysis and RTARG; COREQ and SRQR reporting standards.