Can Generative AI Detect Leading Questions in a Research Interview Guide?

Can AI Identify Interview Questions That Influence Participants' Answers?

You have prepared a qualitative interview guide and want participants to describe their experiences in their own words. The questions appear open-ended, and most begin with "How," "What," or "Can you describe."

One question reads: "How has generative AI improved your teaching effectiveness?" Another asks: "What challenges did you experience when your university restricted AI use?"

Both invite extended answers, but they also contain assumptions. The first presupposes improvement, while the second assumes that institutional restrictions existed and caused difficulties.

Generative AI can review an interview guide and identify such wording. The more difficult task is distinguishing genuinely leading questions from legitimate questions that focus on a particular experience, concept, or research problem.

Yes, AI Can Flag Leading Questions, but Neutrality Requires Context

How AI Can Help Recognize Leading Language in Qualitative Interviews

What Makes a Research Interview Question Leading?

A leading interview question encourages or presupposes a particular response. It may communicate the interviewer's preferred interpretation, assume that an event occurred, or frame an experience in favorable or unfavorable terms before participants have described it themselves.

Consider the question: "Why do teachers resist using generative AI?"

The wording assumes that teachers resist AI and invites participants to explain that presumed resistance. Yet some teachers may use AI enthusiastically, others may be undecided, and some may avoid it for reasons that they would not characterize as resistance.

A less presumptive question might be: "Can you describe how you have approached decisions about using generative AI in your teaching?"

The revised question invites different experiences without requiring participants to accept a particular characterization.

In qualitative interviewing, this matters because the questions help shape which experiences participants consider relevant and how they describe them.

Open-Ended Questions Can Still Be Leading

Researchers sometimes assume that a question is suitable for qualitative inquiry simply because it cannot be answered with yes or no.

However, "How did AI make your teaching more efficient?" is open-ended while presupposing a positive outcome. Similarly, "What problems did your department's poor AI policy create?" introduces a negative evaluation before participants have described the policy.

These qualities are related but not identical. AI can assist in generating open-ended interview questions, but researchers should evaluate the assumptions embedded in each question.

What Types of Leading Questions Can AI Identify?

AI can examine interview questions for several recognizable forms of suggestive framing. Its observations are most useful when it identifies the specific words or assumptions responsible for a concern.

These examples illustrate potential wording problems rather than demonstrating that particular participants would necessarily be influenced by them.

The actual effect of question wording may depend on the preceding conversation, the interviewer's delivery, and the participant's interpretation.

Leading Questions and Necessary Research Focus Are Not the Same

Not every question that directs attention toward a particular topic is methodologically inappropriate.

Suppose a study investigates instructors who have formally reported concerns about institutional AI policies. Asking participants to describe those concerns may be entirely appropriate because the experience forms part of the study's eligibility criteria.

Likewise, an interview about a documented institutional policy may legitimately ask how participants experienced its implementation.

The distinction is whether the question explores an established or participant-reported circumstance or introduces an assumption that has not been justified.

AI may incorrectly flag focused questions as leading when it lacks information about the study design and participant recruitment.

Researchers should therefore supply relevant contextual information before asking AI to assess neutrality.

Can AI Identify Hidden Presuppositions?

Some leading questions contain assumptions that are less obvious than favorable or unfavorable adjectives.

Consider: "When did you realize that AI was affecting your professional judgment?"

This assumes that AI affected the participant's judgment and that the participant eventually recognized the effect.

A more open question might ask: "Have your experiences with generative AI influenced how you make professional decisions, if at all?" If the participant describes an influence, the interviewer can invite a fuller account.

AI can help identify such presuppositions by examining what must be true for the question to make sense as written.

However, the researcher should distinguish assumptions that are genuinely unsupported from circumstances already established during the interview.

Why the Same Question Can Be Appropriate Later in an Interview

A question that would be leading at the beginning of an interview may become an appropriate follow-up after a participant introduces the relevant experience.

For example, asking "How did that restriction affect your teaching?" may be inappropriate if no restriction has been established.

However, if the participant has just described a specific institutional restriction, the question can serve as a reasonable probe.

This is an important limitation of reviewing isolated questions. AI may not know what participants have already said or how the conversation has developed.

Researchers should evaluate leading language in relation to the question's position and function within the interview.

How Should You Prompt AI to Review Leading Interview Questions?

A useful prompt provides the research purpose, participant characteristics, interview context, and complete question list. It should ask the model to identify the exact assumption or wording that may influence an answer.

This prompt encourages a contextual review rather than a mechanical search for positive or negative words.

AI can assist in developing an interview guide, but its assessment should remain subject to researcher judgment.

Do Not Replace Leading Questions With Questions That Are Too Vague

Removing suggestive language can sometimes produce questions so broad that participants struggle to understand what they are being asked.

For example, AI might revise "How did institutional AI restrictions affect your teaching?" into "What are your thoughts about institutions?"

The revision removes the assumption about restrictions but also abandons the phenomenon the researcher intends to explore.

A more useful formulation could be: "What institutional policies or expectations, if any, have influenced your decisions about using generative AI in teaching?"

This retains the substantive focus while allowing participants to report that no relevant policies or expectations influenced them.

The goal is not maximum generality. It is sufficient openness within a clearly defined research inquiry.

Leading Probes Can Be More Subtle Than Leading Main Questions

Researchers may carefully prepare neutral main questions but introduce assumptions during follow-up discussion.

Suppose a participant says, "I decided not to use AI for grading."

A leading probe would be: "Was that because you did not trust the accuracy of AI?"

Unless the participant has already mentioned accuracy concerns, the probe introduces one possible explanation and may make it more salient than others.

A more open follow-up is: "What considerations influenced that decision?"

AI can help prepare interview probes, but their suitability depends on what participants actually say.

DiCicco-Bloom and Crabtree (2006) emphasize the conversational and responsive nature of qualitative interviewing, including the value of following participants' accounts rather than rigidly imposing prepared questions.

Can AI Detect Leading Questions Caused by Interview Sequence?

Question order can influence how participants interpret later questions.

For example, an interview guide might begin with several questions about AI-related risks and misconduct before asking participants to describe their overall experiences with the technology.

Even if the final question is neutrally worded, the preceding discussion may make particular concerns more salient.

AI can examine the full guide and flag possible framing effects, but it cannot establish how actual participants will respond to the sequence.

Researchers should consider the sequence of interview questions alongside their individual wording.

Why Pilot Interviews Still Matter

AI may identify obvious assumptions, but it cannot establish whether actual participants experience a question as leading, confusing, accusatory, or appropriately focused.

Pilot interviewing allows researchers to observe how questions function during interaction and whether participants interpret them as intended.

Castillo-Montoya's (2016) Interview Protocol Refinement framework includes feedback and piloting as part of interview protocol development. Kallio and colleagues (2016) likewise emphasize pilot testing in the development of semi-structured interview guides.

AI-assisted review can help prepare the guide for these procedures, but it cannot replace pilot testing with actual participants.

Using AI to Remove Assumptions From an Interview Question

The revised question preserves the institutional focus while allowing participants to describe positive, negative, mixed, or absent influences.

Its suitability still depends on the study's context and the way it is asked during the interview. Neutral wording should support inquiry rather than eliminate the researcher's substantive focus.

Common Mistakes When Using AI to Detect Leading Interview Questions

Use AI to Expose Assumptions, Not Remove the Research Focus

When reviewing interview questions, ask AI to identify what each question assumes about participants' experiences, motives, or circumstances. Then determine whether those assumptions are supported by the study design or the preceding conversation.

Do not automatically accept every suggested revision. A question can be focused without being improperly leading, and an overly general replacement may no longer address the research inquiry.

Review the complete interview guide, including the planned sequence and optional probes. Then examine its functioning through appropriate feedback and pilot interviewing.

AI can help researchers recognize potential sources of influence, but the interviewer remains responsible for how questions are delivered and how participants' answers are explored.

Before Accepting AI-Suggested Revisions to Leading Interview Questions

Frequently Asked Questions About Leading Qualitative Interview Questions

AI Can Detect Leading Wording, but Researchers Must Judge the Context

How to Cite This Guide

This guide is intended to be read, shared, and used in research, teaching, and academic work. If you draw on its ideas, explanations, or other content, please acknowledge the source by citing the guide. Doing so gives appropriate credit and helps your readers locate the original resource.