Should Research Participants Be Told That AI Will Be Used in the Study?

If You Use AI in Research, Do Your Participants Have a Right to Know?

You are conducting a qualitative study involving university instructors. Participants have agreed to share their experiences through semi-structured interviews, and the consent form explains that their responses will be analyzed by the research team.

After completing the interviews, you decide to use generative AI to help organize transcripts and identify preliminary patterns. The system will not make final analytical decisions, and you intend to verify its suggestions against the original data.

However, you realize that the participant information sheet never mentioned AI. The consent form referred only to research analysis and confidentiality.

You wonder whether participants need to know that an AI system will process their responses. After all, researchers do not ordinarily explain every software application used during analysis.

But generative AI may involve external data processing, automated interpretation, and additional organizations handling research information. Those circumstances may matter to participants in ways that ordinary word-processing software does not.

Should researchers disclose AI use simply because AI is involved, or should disclosure depend on what the technology actually does within the study?

Participants Should Be Informed When AI Use Is Material to Their Participation

When and Why Researchers Should Disclose Generative AI Use to Participants

Informed Consent Is About Understanding the Research, Not Just Signing a Form

Informed consent involves providing prospective participants with information relevant to deciding whether to participate in a study.

Depending on the governing framework, this may include the research purpose, procedures, foreseeable risks, anticipated benefits, confidentiality arrangements, and the voluntary nature of participation.

The Belmont Report identifies respect for persons as a foundational principle and emphasizes the importance of informed, voluntary participation.

Under the United States Common Rule, 45 CFR 46.116 establishes general requirements for informed consent in covered research, including information that a reasonable person would want to have to make an informed decision.

These requirements do not establish a universal rule that every mention of generative AI must appear in every consent form. Rather, they provide a basis for considering whether AI involvement is relevant to participants' understanding of the study.

Researchers should follow the applicable institutional and jurisdictional requirements rather than assume that one consent framework governs all research.

Not Every Use of AI Has the Same Disclosure Implications

Generative AI can assist with research preparation, data collection, analysis, and reporting. These activities may have substantially different implications for participants.

These examples illustrate possible disclosure considerations rather than universal legal classifications.

The researcher should identify what AI actually does and whether that activity changes information participants reasonably need before agreeing to the study.

What Makes AI Use Material to Informed Consent?

Material information is information relevant to an informed decision about participation under the applicable ethical or legal framework.

For example, participants may reasonably consider it important that their interview recordings will be processed by an external AI provider rather than exclusively by the research team.

They may also want to know whether an AI chatbot will conduct the interview, whether their responses will be used to generate additional personal classifications, or whether the processing introduces new confidentiality risks.

However, materiality should not be determined solely by whether a technology is unfamiliar or controversial.

A minor AI-assisted administrative task may have little relevance to participants, while an ordinary-looking automated service may substantially change how their information is processed.

Researchers should evaluate the actual procedures and consequences rather than assume that the word "AI" automatically establishes a disclosure requirement.

Disclosure to an Ethics Committee Is Not the Same as Disclosure to Participants

Researchers may accurately describe AI use in a research ethics application without necessarily providing participants with all the information they need.

For example, an ethics committee may review a detailed description of an AI provider's contractual and security arrangements.

Participants may instead need an understandable explanation of whether their responses will be processed by an external service, what the processing will accomplish, and what safeguards apply.

The two disclosures serve different purposes and may require different levels of technical detail.

Researchers should distinguish describing AI in an ethics application from explaining relevant AI involvement to participants.

Does AI Processing Mean Participants Are Sharing Data With Another Organization?

When research information is submitted to an externally operated AI service, another organization may become involved in processing the data.

Whether that organization is legally classified as a processor, controller, recipient, or another entity depends on the applicable framework and actual arrangements.

For example, a researcher may use an external service to transcribe interviews or summarize qualitative responses. The information may be transmitted to infrastructure outside the university's original research environment.

However, external processing does not necessarily mean the data become public or that the provider is permitted to use them for unrelated purposes.

Researchers should verify the service's contractual conditions, retention arrangements, access restrictions, and permitted uses.

These details may be relevant to what participants should be told.

The distinction is explored further in whether an AI provider becomes a recipient of research data.

Should Participants Be Told That AI Will Analyze Their Interviews?

When generative AI will materially contribute to processing participant interviews, researchers should evaluate whether this involvement needs to be explained in the consent materials.

For example, a study may propose using AI to generate preliminary codes from interview transcripts while researchers retain responsibility for interpreting the findings.

Participants may consider it relevant that an automated system will process their responses, particularly when an external provider is involved.

The researcher should explain the system's actual role without suggesting that it independently understands participants' experiences or determines the final findings.

However, disclosure should not imply that AI analysis is necessarily prohibited or that every use requires separate consent.

The applicable requirements depend on the approved protocol, participant information, data processing arrangements, and institutional guidance.

Should Participants Know if an AI Chatbot Will Interview Them?

When AI interacts directly with participants, its involvement becomes part of the research experience.

For example, an educational technology study might use a generative AI chatbot to ask interview questions or provide individualized feedback during an intervention.

Participants should be given appropriate information about the nature of that interaction under the applicable consent and ethics requirements.

Relevant information may include whether they are communicating with an AI system, what the system is expected to do, whether its responses may vary, and how their information will be recorded or processed.

Researchers should also consider how unexpected AI-generated responses will be managed.

Merely describing the chatbot as an automated research assistant may be insufficient if that description conceals material features of the interaction.

Can Researchers Simply State That AI Will Be Used?

A general statement such as "This study will use artificial intelligence" may be too vague to support meaningful understanding.

Participants may not know whether AI will generate interview questions, process their recordings, analyze transcripts, or produce classifications about them.

A more useful explanation identifies the relevant function and what information the system will receive.

For example, a researcher might explain that selected interview excerpts will be processed using an institutionally authorized AI system to assist with preliminary organization, while researchers will review the original transcripts and make final analytical decisions.

Such a statement should be used only when it accurately reflects the proposed workflow and verified safeguards.

The amount of detail appropriate for participant materials is considered further in how much participants should be told about AI processing.

Do Researchers Need to Explain How the AI Model Works?

Meaningful disclosure does not necessarily require participants to understand the mathematical architecture or training procedures of a large language model.

For example, participants may need to know that an external AI service will process interview transcripts, but they may not need an explanation of transformer architectures or model parameters.

Technical detail becomes relevant when it affects the information needed to understand the research, its risks, or the participant's choices.

Researchers should avoid overwhelming participants with technical terminology that obscures the practical implications of the processing.

At the same time, oversimplification should not conceal uncertainty or important limitations.

The objective is understandable and accurate communication rather than technical completeness for its own sake.

Should Researchers Name the Specific AI Provider?

Whether the provider must be identified by name depends on the applicable disclosure requirements and the significance of the provider's role.

For example, participants may reasonably consider it important whether their information will be processed by an external commercial provider or remain within an institutionally managed environment.

In some circumstances, the specific provider identity may be required or materially relevant to the processing description.

In others, an accurate explanation of the processing arrangement may be sufficient under the applicable institutional requirements.

Researchers should not assume that naming ChatGPT alone provides adequate disclosure. The same provider may offer different products and contractual arrangements.

The separate question of whether the AI provider should be named in the consent form requires attention to the actual service arrangement.

What About Data Retention and Processing Locations?

Participants may need information about how long their data will be retained and whether the processing involves organizations or infrastructure outside the original research institution.

The required details depend on applicable data-protection law, institutional policy, and the nature of the processing.

For example, an external AI provider may retain certain submitted information or logs under specified conditions. Researchers should verify those conditions rather than assume that deleting a chat immediately removes every copy.

Similarly, processing may involve international data transfers that require additional legal or institutional assessment.

Researchers should not promise that participant information remains in a particular country or is deleted immediately unless the applicable arrangement supports that assurance.

These concerns are addressed separately in disclosing AI processing locations and disclosing AI provider data retention.

Does Institutional Approval Mean Participants Do Not Need to Know?

No. Institutional authorization and participant information serve different purposes.

A university may approve a particular AI service for specified categories of research information. However, that authorization does not automatically determine what participants must be told about the proposed processing.

For example, an institutionally authorized enterprise AI service may provide contractual safeguards while still processing participant information through an external provider.

Whether that arrangement requires participant disclosure depends on the study and applicable requirements.

Researchers should avoid assuming that technical approval eliminates the relevance of participant expectations.

The distinction between formal permission and what participants reasonably expected remains important when assessing the ethical acceptability of AI-assisted research.

What Does the Philippine Data Privacy Act Mean for Disclosure?

In the Philippines, Republic Act No. 10173, or the Data Privacy Act of 2012, establishes requirements concerning the processing of personal information.

Its transparency principle and related provisions are relevant to informing data subjects about the nature, purpose, and extent of personal information processing.

Section 16 also identifies rights of data subjects, including rights to be informed about specified aspects of processing, subject to the Act's provisions and applicable exceptions.

For example, a Philippine university researcher planning to submit identifiable interview information to an external AI service should consider whether the participant information materials accurately explain the relevant processing arrangements.

However, the Act should not be interpreted as establishing a universal requirement to identify every software tool by name in every research consent form.

Researchers should examine the applicable legal requirements, National Privacy Commission guidance, and institutional ethics procedures.

Research participation consent and the legal basis for processing personal information should also be distinguished rather than treated as automatically identical.

How Can Researchers Explain AI Use in a Consent Form?

A useful explanation should describe the actual role of AI, the information involved, and the relevant safeguards without making unsupported assurances.

This wording is illustrative. Researchers should not use it unless the authorization, safeguards, and described analytical procedures have actually been established.

Where an external provider is involved, the institution may require additional information about the provider, data retention, processing location, or participant rights.

The final language should be reviewed according to the applicable ethics and data-protection requirements.

Can AI Help Researchers Draft Participant Disclosure Language?

Generative AI can assist researchers in preparing understandable descriptions of proposed research procedures.

However, the model should not be allowed to invent safeguards, consent conditions, provider policies, or institutional authorization.

This prompt can assist with drafting, but the researcher must verify the resulting statement against the approved protocol and actual processing arrangement.

AI-generated consent language should not be treated as independently reviewed or legally sufficient merely because it sounds clear and professional.

Should Participants Be Allowed to Refuse AI Processing?

Some participants may be comfortable participating in a study but prefer that their information not be processed through an AI system.

Whether researchers must provide an AI-specific opt-out depends on the study design, applicable consent and data-protection requirements, and institutional determination.

For example, a research team may be able to offer researcher-led analysis as an alternative to AI-assisted analysis.

However, such an option should be offered only when it can genuinely be implemented without misleading participants.

In other studies, AI processing may be integral to the intervention being investigated, making participation without AI incompatible with the research design.

Researchers should distinguish voluntary participation in the study from the availability of specific choices about processing methods.

The separate question of whether participants should be able to opt out of AI processing requires consideration of both ethical expectations and operational feasibility.

What If AI Is Introduced After Participants Have Already Consented?

Researchers may decide to use generative AI after recruitment or data collection has begun.

For example, participants may have consented to interviews analyzed by the research team, while the researcher later proposes submitting transcripts to an external AI provider.

The researcher should not assume that the original consent automatically covers every new processing arrangement.

However, additional consent is not necessarily required for every technological modification.

The appropriate procedure depends on whether the new activity changes the approved purpose, participant information, confidentiality arrangements, or risks.

Researchers should consult the applicable institutional process before implementing consequential changes.

Questions about AI introduced after participants have consented and whether reconsent is required should be evaluated according to the actual modification.

Can Broad Consent Cover AI Use?

Some studies obtain broad consent for specified categories of future research or data use.

Whether that consent covers a particular AI-assisted analysis depends on its scope, the information originally provided, the proposed processing, and the governing framework.

Broad consent should not be interpreted as unrestricted permission to use participant information for any future technological application.

At the same time, the presence of AI does not necessarily invalidate an otherwise authorized future analysis.

Researchers should examine whether the proposed activity falls within the agreed purposes and applicable safeguards.

It is important to distinguish broad consent for future AI analysis from a general assumption that consent to data sharing automatically authorizes all forms of AI processing.

What If Participants Ask Questions About AI?

Participants may ask whether their responses will be used to train AI models, whether a provider can access their transcripts, or whether they can withdraw their information after processing.

Researchers should answer using verified information about the actual service arrangement.

They should not promise that data will never be retained, that no personnel can access the information, or that deletion is immediate unless those claims are supported by the applicable conditions.

Where a question concerns institutional authorization or legal rights beyond the researcher's expertise, the appropriate institutional office should be consulted.

Researchers should also be willing to acknowledge uncertainty rather than provide reassuring but unsupported answers.

Meaningful informed consent depends on accurate communication, including when the available information has limitations.

Does Disclosing AI Use Automatically Make the Research Ethical?

No. Disclosure is one component of responsible research, but it does not independently establish that the proposed processing is necessary, lawful, methodologically sound, or appropriately protected.

For example, telling participants that their transcripts will be uploaded to an unauthorized external service does not automatically make that transfer acceptable.

Similarly, disclosing that AI will generate psychological classifications does not establish that those classifications are valid or justified.

Researchers should separately evaluate authorization, participant welfare, data protection, methodological validity, and the balance of risks and benefits.

Sometimes the appropriate decision may be not to use the proposed AI system, even when its involvement could be fully disclosed.

Explaining AI-Assisted Interview Analysis Before Participants Consent

The researcher does not need to explain every technical feature of the AI model. However, the information should accurately communicate the processing arrangements relevant to participants' decisions.

The example illustrates how AI disclosure can become part of meaningful informed consent rather than a generic statement added merely because the study uses modern technology.

Common Misconceptions About Disclosing AI Use to Research Participants

Explain AI When It Changes What Participants Need to Understand

Before preparing participant information materials, identify the actual role of AI in the study.

Determine whether the technology affects participant interactions, data processing, confidentiality, or the information generated about participants. Then establish what must be disclosed under the applicable institutional and legal requirements.

Use language that participants can understand, and verify every statement about the provider and safeguards.

The objective is not to make consent forms longer simply because AI is involved. It is to ensure that participants are not left with a materially inaccurate understanding of what will happen to their information.

Before Finalizing Participant Disclosure About Generative AI

Frequently Asked Questions About Disclosing AI Use to Research Participants

Participants Should Understand the Research They Are Agreeing To, Including Material AI Involvement

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.