The Identity Question Nobody Is Asking

Grodal opens the conversation with a question that reframes the entire debate: "Are we qualitative researchers because we engage in open coding, because we identify themes in data using Atlas or NVivo? Or are we qualitative researchers because we produce new theories that move our understanding forward?"

This distinction matters enormously for new scholars. A lot of the anxiety around AI in qualitative research is really anxiety about identity. If what makes someone a qualitative researcher is the act of manually coding data, then AI feels like a direct threat. But if what makes someone a qualitative researcher is the ability to think critically, to sit with complexity, to develop theory from the mess of human experience, then AI is a tool. A powerful one, but a tool.

New scholars are entering a field at a moment when practices are changing faster than the textbooks. That is uncomfortable. It also means you are not stuck defending old methods the way some senior scholars are. You have the advantage of building your identity around what qualitative research produces, not just how it has always been done.

The Fear Is Real, But It Is Not New

One of Schildt's most useful points in the conversation is the historical parallel. When computers first appeared in the 1940s, people were genuinely unsettled because calculation was considered a uniquely human trait. It seems almost absurd now, but at the time it was a serious existential question.

The field is doing the same thing with AI today. Because AI can write, summarize, and identify patterns in text, researchers are asking whether it is replacing the human researcher. That anxiety is driving a lot of the knee-jerk bans in journals and universities.

This pattern is well documented. Existing studies on AI adoption in qualitative research remain scarce and limited in scope, with most focusing on descriptive analyses rather than in-depth examinations of methodological and epistemological implications. In other words, the field is reacting before it fully understands what it is reacting to. New scholars have the opportunity to engage with AI more thoughtfully than the first wave of panic allowed for.

Automation vs Augmentation: The Line That Changes Everything

This is where the conversation is at its most valuable. Grodal draws a clear distinction between two ways of using AI, and the difference between them is not subtle.

Automation is when you hand your data to AI and accept what comes back. You let it code, summarize, and interpret, and you report those outputs as findings. Grodal is direct about the risk: "In qualitative research, the legitimacy of our work lies with the human interpreter. We cannot just automate our work with AI and put the legitimacy of our work inside of the tool."

Augmentation is different. It is when you use AI to think better, to challenge your assumptions, surface things you might have missed, and push your analysis further than you could alone. The legitimacy still sits with the researcher. The tool helps get there.

Recent advancements in AI, particularly large language models, have shown potential in qualitative analysis by supporting tasks like labeling and collaborative coding. These models tend to excel in deductive analysis but may face limitations in inductive reasoning and maintaining code diversity. This is exactly why human oversight is not optional. It is what makes the work qualitative in the first place.

One of the more interesting moments in the conversation is when Schildt talks about confirmation bias. He points out that AI might actually help researchers revisit interviews they had mentally written off, ones they assumed had nothing interesting to offer. That is augmentation at its best: not replacing judgment, but expanding it.

What AI Actually Opens Up for New Scholars

AI does not just make existing research faster. It makes previously impossible research possible.

Grodal gives a compelling example. If you want to study institutional change across decades and across continents, you need enormous datasets. Traditionally, that kind of scope was simply out of reach for most qualitative researchers. Too much data, not enough time, too few resources. AI changes that equation.

The cost of using powerful AI has dropped significantly. Querying a model with GPT-3.5 level performance fell from about $20 per million tokens in late 2022 to just a few cents by mid-2024. For new scholars without large research budgets, this matters. The tools that were once available only to well-funded labs are now accessible to a PhD student working alone.

Think about what that means for the questions you can ask. Longitudinal studies. Cross-cultural comparisons. Massive archival datasets spanning decades. These are no longer out of reach just because you do not have a team of twenty.

As Grodal put it: "AI opens up for us to ask new questions and study things we couldn't before."

The Deskilling Trap and How to Avoid It

This is the part of the conversation that should make every new scholar pay close attention. A junior researcher in the audience raised a concern that comes up constantly: using AI so much that you start losing your own critical faculties. Your ability to theorize, to write, to think through complexity on your own.

It is a real risk, and Grodal does not dismiss it. Her advice is blunt: "You need to kind of force yourself all the time to read through what is the AI telling you and do you think that's right? And in qualitative research, always go back to the data yourself."

Researchers working in this area have found that AI can expedite coding and uncover new insights, but only if researchers remain transparent, critically evaluate all outputs, and preserve human reflexivity. That last phrase, human reflexivity, is the thing AI cannot replicate. It is the ability to sit with uncertainty, to change your mind, to be changed by your data.

Practically speaking, here is what that looks like:

• Code a portion of your data yourself before you let AI anywhere near it. Build your own interpretation first.

• When AI surfaces a theme or a quote, go back to the original source every time.

• Treat AI outputs the way you would treat a suggestion from a smart but untrained colleague: interesting, worth investigating, not automatically trustworthy.

• Use AI for specific, bounded tasks: summarizing a document, flagging contradictions in your data, generating a first-pass overview of a large dataset. Not for drawing conclusions.

The Journal Question

One of the more honest exchanges in the conversation is about publishing. Grodal admits she has not used AI in any paper she has submitted to journals, not because she does not think it is valuable, but because journal policies currently do not allow it and she wants to get published. Schildt adds that some journals, like AMJ, are beginning to open up, allowing AI to be disclosed as part of the methodology. But he is candid: "There are a lot of qualitative researchers who are very much categorically against AI."

A paper that discloses AI use might face a reviewer who rejects it on principle, regardless of the quality of the work.

For new scholars, this is a practical reality to navigate carefully. For now, the safest approach is to use AI in your research process for exploration, for analysis, for making sense of large datasets, but to be thoughtful about what you disclose and where. Check your target journal's policy before you submit. If you do use AI, disclose it in your methods section honestly.

This will change. The community is actively working through what responsible AI use looks like in qualitative research, and the conversation Grodal and Schildt are having is part of that process. The scholars publishing on this today are helping shape the norms that will govern the field in five years.

A Word on Security and Confidentiality

Schildt addresses this directly in the conversation, and it is worth paying close attention to, especially for anyone working with sensitive interview data.

Not all AI tools are created equal when it comes to data privacy. Schildt is direct: he would not trust a standard individual subscription to ChatGPT or Claude with personal research data. The reason is not the quality of the tools. It is compliance. Most individual subscriptions are not GDPR compliant, which is a serious issue for researchers working with human subjects data in Europe and increasingly elsewhere. For researchers in healthcare or working with clinical populations, HIPAA compliance adds another layer of complexity.

The safer options Schildt points to are corporate subscriptions to tools like Google's NotebookLM and Microsoft Copilot, which come with data processing agreements and are already trusted by law firms and government institutions.

The practical takeaway for new scholars is simple: before uploading any interview transcript or sensitive data to an AI tool, check two things. First, does your university have a data processing agreement with that provider? Second, are you on a corporate or institutional subscription, or just a personal one? If you cannot answer both of those questions confidently, do not upload the data.

Qualtranscribe's Instant Draft and Smart Insights are built with GDPR and HIPAA compliance from the ground up, including regional data storage, so sensitive research data stays protected while AI does the heavy lifting. Recordings are never used to train AI models under any plan. For more on this topic, see The Ethics of AI in Academic Research Transcription.

A Practical Starting Point

If you are a new scholar trying to figure out where to begin, here is a simple framework drawn from the conversation and from how researchers use these tools in practice:

Start small. If you have a large dataset, do not feed all of it to AI at once. Take a small sample, read it yourself, develop your own initial sense of what is interesting. Then use AI to summarize the broader dataset and compare.

Be specific. Qualtranscribe's Instant Draft and Smart Insights are built for exactly this: helping researchers transcribe interviews quickly and surface insights from their data without replacing the critical thinking that makes qualitative work meaningful.

Stay in the data. Every time AI points you somewhere interesting, go there yourself. Read the context. Ask whether the AI's interpretation holds up. This is not extra work. It is the real work.

Document your process. As norms around AI disclosure evolve, having a clear record of how and where you used AI will protect you and strengthen your methodology section.

The microwave analogy Schildt uses in the conversation captures the moment well. When microwaves first appeared, people thought they would replace ovens entirely. They did not. The field figured out what they are actually good for and that is mostly what they are used for now. AI in qualitative research is at that same stage: the field is still working out what it is actually good for and what it will never be able to replace. The researchers who will shape that conversation are the ones willing to engage seriously, critically, and honestly. New scholars are uniquely positioned to lead that work, entering the field without decades of entrenched practice to defend.

The transcript of the Using AI in Qualitative Research webinar featured in this post was generated using Qualtranscribe's AI Transcription. Download the full transcript here. Watch the full conversation on YouTube here.

Want to transcribe your own research interviews? Try Qualtranscribe's AI transcription free today.

Frequently Asked Questions

What is the difference between automation and augmentation in AI qualitative research? Automation means handing your data to AI and reporting its outputs as findings. The legitimacy of the work shifts to the tool. Augmentation means using AI to extend your thinking while you maintain interpretive control. The legitimacy stays with you. Grodal's distinction between the two is the most useful framework for new scholars deciding how to integrate AI into their research practice.

Is it ethical to use AI for qualitative research transcription? Yes, if the platform meets your compliance requirements, does not train on your data, and your consent language and IRB protocol cover third-party AI processing. Generic consumer AI tools frequently fail on one or more of these criteria. Purpose-built research tools like Qualtranscribe Instant Draft are designed to meet them. See The Ethics of AI in Academic Research Transcription for a full breakdown.

Do I need to disclose AI use in my journal submissions? Check your target journal's policy before submitting. Policies vary significantly and are changing rapidly. Some journals now permit AI use with disclosure in the methods section. Others prohibit it entirely. As Schildt notes, a paper disclosing AI use may still face reviewers who reject it on principle regardless of quality.

Is my individual ChatGPT or Claude subscription safe for research data? Schildt addresses this directly: individual subscriptions to standard consumer AI tools are typically not GDPR compliant and should not be used for sensitive research data. Use institutional or corporate subscriptions with documented data processing agreements, or purpose-built research platforms with explicit compliance documentation.

What AI tools are safe for HIPAA-governed research? Tools that are HIPAA compliant, can sign a Business Associate Agreement, use encrypted file handling, and have a documented policy against using your data for AI training. Qualtranscribe Instant Draft meets all of these from the Pro plan upward. Most consumer AI transcription tools do not. See Data Security in Academic Research Transcription for what to check.

Will AI replace qualitative researchers? Grodal and Schildt's answer is nuanced and worth hearing in full. The short version: AI can automate tasks that qualitative researchers do, but the interpretive, theoretical, and reflexive dimensions of qualitative work sit with the human researcher and are not replicable by current AI. The risk is not replacement. It is deskilling through overreliance. Protect your critical faculties by staying in the data and treating AI outputs as suggestions rather than findings.

What is Smart Insights and how does it help qualitative researchers? Smart Insights is Qualtranscribe's built-in analysis layer on Instant Draft transcripts. It automatically surfaces recurring themes, extracts key quotes with speaker attribution and timestamps, identifies sentiment patterns, and generates research-ready summaries. It is designed for augmentation: giving researchers a faster starting point for analysis while keeping interpretive control with the researcher.

References

• Grodal, S., & Schildt, H. (2025). Qualitative research with artificial intelligence: The threat of automated uses and the augmented alternative. Working paper.

• Lumivero. (2025). The state of AI in qualitative research in 2025. https://lumivero.com/resources/blog/state-of-ai-in-qualitative-research/

• Chatzichristos, G. (2025). Qualitative research in the era of AI: A return to positivism or a new paradigm? International Journal of Qualitative Methods.

• Nicmanis, M., & Spurrier, H. (2025). Getting started with artificial intelligence assisted qualitative analysis. International Journal of Qualitative Methods.

Related Reading

• Qualitative Research in 2026: Growth, Remote Studies, AI, and the Transcription Gap

• Can I Use AI Transcription for IRB-Approved Research?

• The Ethics of AI in Academic Research Transcription

• Understanding Artificial Intelligence Through the IRB Lens

• From Transcripts to Themes: A Practical Guide to Thematic Analysis for Researchers