Can Generative AI Help Assess Sample Adequacy in Qualitative Research?

Can AI Tell You Whether 15 Interviews Are Enough for Your Qualitative Study?

You are planning a qualitative investigation of university instructors' experiences using generative AI in assessment. You ask an AI chatbot how many participants you should recruit.

It recommends 15 interviews, explaining that qualitative studies commonly use between 10 and 20 participants and that saturation will probably occur within that range.

The recommendation sounds practical. But what if your participants come from substantially different disciplines? What if some have extensive experience with AI assessment while others have only recently encountered the technology?

And what if your study uses grounded theory rather than a descriptive interview approach?

Qualitative sample adequacy cannot be established simply by selecting a familiar number of participants. It depends on the purpose of the inquiry, the sampling strategy, the quality and relevance of the material, and the analytical claims the researcher intends to develop.

AI Can Help Evaluate Qualitative Sample Adequacy, but It Cannot Certify Sufficiency From Participant Counts Alone

How AI Can Help Evaluate Qualitative Sampling and Sample Adequacy

What Does Sample Adequacy Mean in Qualitative Research?

Qualitative sample adequacy concerns whether the participants, cases, observations, or other materials provide sufficient and relevant information for the study's purpose and analytical approach.

Unlike conventional probability-based surveys, many qualitative studies do not aim to estimate population parameters within a specified margin of error.

Instead, researchers may seek to understand experiences, interpret meanings, examine social processes, or develop conceptual explanations.

The adequacy of a sample therefore depends on what the study intends to accomplish and how the evidence is analyzed.

Patton (2015) discusses purposeful sampling as a means of selecting information-rich cases relevant to qualitative inquiry.

AI should begin by identifying the methodological purpose rather than immediately recommending a numerical target.

Why Is There No Universal Qualitative Sample Size?

Different qualitative traditions require different forms of evidence and analytical development.

For example, a focused interview study involving participants with highly specific experience may require fewer interviews than a study seeking to understand substantial variation across institutions, professional roles, and educational settings.

A grounded theory study may use theoretical sampling to develop emerging categories, while a case study may draw on multiple data sources within a bounded case.

Consequently, the same number of participants may be adequate for one study and insufficient for another.

AI-generated recommendations such as "15 participants are enough" should therefore be treated as provisional planning suggestions rather than methodological conclusions.

What Factors Influence Qualitative Sample Adequacy?

These factors interact. A narrowly defined study with highly specific participants and detailed interviews may have different requirements from a broad study involving heterogeneous participants and limited interview depth.

What Is Information Power?

Information power is an approach proposed by Malterud, Siersma, and Guassora (2016) for considering sample size in qualitative interview studies.

The central principle is that the more relevant information a sample provides for the study aim, the fewer participants may be needed, depending on the analytical purpose.

The framework identifies five considerations:

Information power is not a mathematical formula that produces a universally correct sample size.

AI may help researchers examine the five considerations, but their application requires judgment about the actual inquiry and collected material.

Can AI Calculate Information Power as a Numerical Score?

AI may offer to rate information power on a numerical scale or assign points to each consideration.

However, Malterud and colleagues' framework does not establish a universal numerical scoring system that converts the five dimensions into an exact participant requirement.

A score such as 85% information power would therefore require independent methodological justification and should not be presented as a validated measure merely because AI generated it.

A more defensible approach is to document how the study aim, participant specificity, theoretical orientation, dialogue quality, and analytical strategy inform sampling decisions.

The more detailed question of AI-assisted information power assessment should be approached as a qualitative methodological judgment rather than a mechanical calculation.

What Is Data Saturation?

Saturation is a concept used in several qualitative approaches to describe a point at which further data collection no longer contributes sufficiently new information, categories, or analytical insights under the study's specified criteria.

However, saturation is not defined identically across qualitative methodologies.

For example, identifying no additional descriptive codes is not necessarily equivalent to developing a theoretically adequate explanation of a social process.

Hennink, Kaiser, and Marconi (2017) distinguish code saturation from meaning saturation, emphasizing that identifying issues and understanding them in depth are different analytical achievements.

AI should therefore clarify what form of saturation the study intends to assess.

Is Saturation Appropriate for Every Qualitative Study?

No. Some qualitative traditions use saturation as an important sampling and analytical consideration, while others question its appropriateness.

For example, Braun and Clarke (2021) argue that saturation is not conceptually consistent with reflexive thematic analysis as they understand it, particularly when themes are treated as interpretive constructions rather than discoveries that can be exhausted.

Consequently, AI should not recommend saturation automatically whenever a researcher mentions qualitative interviews or thematic analysis.

The appropriateness of saturation depends on the methodological tradition and analytical assumptions.

Researchers should distinguish between the general concern for adequate material and the particular logic of saturation.

What Is the Difference Between Code Saturation and Meaning Saturation?

Code saturation concerns the point at which additional data collection produces few or no new codes or issues under the analytical approach being used.

Meaning saturation concerns whether the researcher has developed a sufficiently rich understanding of the identified issues.

A researcher may identify the main categories of instructors' AI assessment experiences relatively early while requiring additional interviews to understand differences across disciplines or institutional conditions.

AI should not equate repeated code labels with adequate conceptual understanding.

Can AI Determine Whether Saturation Has Been Reached?

AI can help organize coding records, compare successive interviews, and identify whether new categories appear under a specified coding framework.

However, it cannot independently certify saturation simply because similar themes recur in the material.

The judgment depends on the analytical purpose, the depth of interpretation, the variation among participants, and the criteria adopted by the researcher.

Automated coding may also overlook contextual distinctions or produce overly broad categories that make the material appear more repetitive than it is.

Researchers should therefore treat AI-assisted saturation assessment as an aid to analysis rather than a substitute for methodological judgment.

How Does Purposive Sampling Affect Adequacy?

Purposive sampling selects participants or cases according to characteristics relevant to the research question.

For example, a study examining instructors' experiences implementing generative AI assessment policies may recruit participants who have direct responsibility for assessment design.

Such participants may provide more relevant information than individuals with little experience of the phenomenon.

However, selecting highly experienced participants may also limit the perspectives represented.

For example, instructors who have successfully implemented AI policies may describe different experiences from those who encountered substantial difficulties.

Researchers should therefore consider whether the sample includes the variation necessary for the study's analytical purpose.

What Is Maximum Variation Sampling?

Maximum variation sampling is a purposive strategy that seeks cases differing across characteristics relevant to the phenomenon.

For example, a researcher may select instructors from different disciplines, institutional types, or levels of AI experience.

The purpose is not necessarily statistical representativeness.

Instead, the researcher may seek to understand how experiences differ across meaningful contexts and whether important patterns appear despite those differences.

However, greater variation can also increase the information required to understand each subgroup or context adequately.

AI should therefore not recommend maximum variation sampling without considering the study's scope and analytical capacity.

How Does Theoretical Sampling Differ From Initial Purposive Sampling?

Theoretical sampling is associated particularly with grounded theory approaches.

It involves selecting additional cases or data sources in response to emerging analytical categories and the need to develop theoretical understanding.

For example, interviews may suggest that institutional assessment policies shape how instructors use generative AI.

The researcher may then seek participants from settings with contrasting policy arrangements to investigate the developing explanation.

This differs from selecting every participant according to a fixed demographic quota established before analysis begins.

AI may help identify gaps in emerging categories, but theoretical sampling decisions should follow the actual analytical development of the study.

Can AI Recommend Sample Sizes Based on Published Qualitative Studies?

AI can summarize sample sizes reported in comparable research and help researchers understand disciplinary practices.

However, the fact that a published study interviewed 12 participants does not establish that 12 participants will be adequate for another investigation.

The studies may differ in their aims, participant characteristics, interview quality, methodological traditions, and analytical strategies.

Hennink and Kaiser (2022) systematically review empirical studies concerning saturation in qualitative research and discuss the conditions under which particular sample ranges have been observed.

Such evidence can inform planning, but it should not be converted into a universal numerical requirement.

AI should distinguish empirical observations from prescriptive rules.

Does a Larger Qualitative Sample Always Improve the Study?

No. Additional participants may contribute new perspectives or strengthen analytical development, but more data also increase the demands of collection, interpretation, and analysis.

A large sample does not automatically compensate for superficial interviews, poorly justified participant selection, or an incoherent analytical approach.

Conversely, a smaller sample may be defensible when the inquiry is narrowly focused and the material supports sufficiently detailed interpretation.

The relevant question is whether additional data would meaningfully contribute to the study's purpose.

Researchers should avoid treating sample size as a numerical indicator of qualitative rigor independent of analytical quality.

Can AI Assess Sample Adequacy Before Data Collection?

AI can help develop a preliminary sampling rationale and identify factors likely to influence the amount of material required.

For example, it may highlight that a study spanning several institutional types and academic disciplines could require broader recruitment than a focused inquiry involving one specialized participant group.

However, the richness and relevance of future interviews cannot be known with certainty before data collection.

Researchers should therefore distinguish a provisional recruitment plan from a final judgment about sample adequacy.

Where the methodology permits, sampling may be reviewed iteratively as analysis develops.

Can AI Assess Sample Adequacy During Data Collection?

AI may assist in organizing recruitment records, comparing participant characteristics, summarizing analytical gaps, and identifying areas requiring further investigation.

For example, it may help researchers examine whether the sample includes instructors from the disciplinary contexts specified in the sampling rationale.

It may also help compare the analytical contribution of successive interviews when appropriate and authorized material is available.

However, the researcher must examine the original data and determine whether the analysis has developed sufficiently.

AI-generated summaries should not replace close engagement with participants' accounts or become the sole basis for deciding that recruitment can stop.

Can AI Assess Sample Adequacy After Data Collection?

AI can help organize a retrospective justification by comparing the achieved sample, participant characteristics, research aim, and analytical findings.

However, researchers should not use AI to manufacture a claim of saturation or information power after collecting an arbitrarily convenient number of interviews.

A retrospective account should accurately describe the sampling decisions made, the evidence obtained, and the limitations that remain.

For example, if the study includes only instructors who voluntarily attended an AI training workshop, the researcher should acknowledge that instructors with different experiences may be underrepresented.

The adequacy argument should reflect what the study actually accomplished rather than what the researcher initially hoped to achieve.

Can AI Evaluate Qualitative Sample Adequacy Without Access to the Data?

Only provisionally. Without the actual interviews, observations, documents, or analytical records, AI cannot assess the richness or contribution of the empirical material.

It may evaluate whether the proposed sampling plan is conceptually coherent and whether the recruitment strategy addresses the study's stated purpose.

However, it cannot determine whether the final data support sufficiently developed interpretations.

Even when AI is provided with transcripts or coding records, its conclusions require researcher verification.

Confidentiality and institutional data governance requirements must also be considered before submitting participant material to an AI system.

Should Qualitative Researchers Report an Exact Sample Size Justification?

Researchers should explain how and why participants or cases were selected and why the achieved material was considered adequate for the intended analysis.

The form of justification depends on the methodological approach.

For example, an information power rationale may discuss study specificity, participant relevance, interview quality, and analytical strategy.

A grounded theory study may explain how theoretical sampling contributed to category development.

A reflexive thematic analysis may justify its dataset through the research purpose, scope, and interpretive depth without claiming saturation.

The justification should be transparent and consistent with the methodological assumptions of the study.

How Can You Conduct an AI-Assisted Qualitative Sample Adequacy Review?

This process supports transparent qualitative sampling decisions without reducing adequacy to a fixed participant count.

When AI Recommends 15 Interviews Without Considering the Study's Scope

The researcher rejects the claim that 15 interviews will necessarily achieve saturation.

Instead, the researcher develops a purposive sampling strategy aligned with the study aim and selects an adequacy rationale appropriate to the chosen qualitative methodology.

If the study uses an approach for which information power is appropriate, the researcher may examine the specificity of the sample, quality of dialogue, and analytical strategy.

If the study uses reflexive thematic analysis, the researcher should avoid imposing a saturation criterion that conflicts with the methodological approach.

The decision: The recruitment target is treated as provisional, while final adequacy is justified through the relationship between the study purpose, selected cases, analytical approach, and empirical material.

The example demonstrates why a familiar interview count cannot replace methodological reasoning.

Common Mistakes When AI Evaluates Qualitative Sample Adequacy

Ask AI Whether the Sample Supports Your Analysis, Not Whether the Number Looks Acceptable

When using AI to evaluate qualitative sample adequacy, begin with the methodological tradition and research aim.

Ask AI to examine the sampling rationale, participant relevance, intended variation, and adequacy criteria appropriate to the selected approach.

A Prompt for Reviewing Qualitative Sample Adequacy

Researchers should then examine whether saturation is appropriate and how it should be assessed, or consider information power where it fits the methodology.

The final justification should explain how the sample contributes to the study's analytical purpose rather than rely on a predetermined number presented as universally sufficient.

Before Accepting an AI Qualitative Sample Adequacy Assessment

Frequently Asked Questions About AI and Qualitative Sample Adequacy

Qualitative Sample Adequacy Is About the Contribution of the Evidence, Not the Count Alone

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.