Does Using Generative AI in Research Automatically Require Ethics Approval?

Do You Need Ethics Approval Every Time You Use ChatGPT in Research?

You are preparing a research study and have started using generative AI to help organize your literature review, refine interview questions, and improve the wording of your methodology.

A colleague suggests that any use of ChatGPT in academic research requires approval from a research ethics committee. Another argues that AI is merely a software tool and therefore falls entirely outside ethics review.

Neither position adequately addresses the range of activities researchers may undertake with AI.

Using AI to suggest a nonconfidential research title is different from uploading identifiable interview transcripts, generating participant classifications, or introducing an AI chatbot that interacts directly with research participants.

The relevant question is not simply whether generative AI appears somewhere in the research process. It is whether the proposed activity falls within applicable ethics review requirements or changes the ethical characteristics of the study.

No, AI Use Alone Does Not Automatically Trigger Ethics Approval

When Generative AI Use May Require Research Ethics Review

Research Ethics Approval Concerns the Activity, Not Merely the Technology

Research ethics committees, institutional review boards, and comparable oversight bodies evaluate research according to applicable ethical and regulatory requirements.

Depending on the jurisdiction and institution, relevant considerations may include participant involvement, recruitment, consent, risks, confidentiality, and the collection or processing of identifiable information.

The United States' Common Rule, for example, defines research involving human subjects through specified interactions, interventions, and uses of identifiable private information or identifiable biospecimens. It does not establish a universal requirement for review merely because a particular software tool is used.

However, the Common Rule is a United States regulatory framework, not a universal rule governing every university or country. Institutional policies may impose additional requirements.

Researchers should therefore examine what AI will actually do within the study and which review framework applies.

Different AI Activities May Have Different Review Implications

The following examples illustrate why a single rule for all AI use would be inadequate.

These are illustrative assessments, not universal classifications. The responsible institutional authority determines the applicable review requirements.

Distinguish Research Ethics Approval From Permission to Use AI

A researcher may have ethics approval for a study without having authorization to use every external AI service.

For example, an approved protocol may permit interviews and thematic analysis, but institutional information security policies may restrict uploading participant records to external platforms.

Conversely, a university may permit employees to use an enterprise AI service without that permission constituting ethics approval for a particular human-participant study.

These processes may overlap, but one does not automatically substitute for the other.

A study may require both an appropriate ethics determination and authorization for the proposed AI processing environment.

Does Using AI With Human-Participant Data Automatically Require a New Application?

Not necessarily. The answer depends on whether the proposed processing is already covered by the approved protocol and relevant institutional arrangements.

For example, a study may have an approved data analysis plan that explicitly includes a particular institutionally managed AI system, defined safeguards, and appropriate participant information.

Using that system within the approved conditions may not require a separate application each time the researcher conducts an analysis.

However, transferring the same data to a new external provider with different retention or access conditions may require further review.

Researchers should not assume that all AI systems are interchangeable simply because they perform similar analytical functions.

Whether an AI provider becomes an additional recipient of research data is particularly relevant when assessing new processing arrangements.

What If AI Is Used Only Before Participant Recruitment?

AI assistance during research preparation may not independently trigger an additional ethics approval requirement.

For example, researchers may use AI to brainstorm potential interview topics or identify wording problems before submitting their protocol.

However, the resulting questions and procedures still need to be evaluated within the research design. An AI-generated interview guide does not become ethically appropriate merely because it was developed before recruitment.

If the planned study requires ethics approval, the relevant participant-facing procedures should be included in the application according to institutional requirements.

Researchers should distinguish assistance with preparing a protocol from the authorization required to implement that protocol.

Can AI-Mediated Participant Interaction Require Ethics Review?

Yes. When an AI system interacts directly with participants, it may become part of the research intervention or data collection procedure.

For example, a study might use a generative AI chatbot to conduct interviews, provide personalized educational feedback, or respond to participants during an experimental learning activity.

The ethical assessment may need to consider what participants are told about the system, whether it can generate unexpected responses, what information it collects, and how researchers monitor or manage potentially harmful interactions.

AI-mediated participation can also introduce concerns involving accessibility, misleading information, unequal treatment, and participants' understanding of whether they are interacting with a human or an AI system.

These considerations should be addressed in the applicable protocol rather than assuming that the technology's automated operation places it outside human-participant research ethics.

What If the AI System Processes Only De-Identified Data?

Removing direct identifiers may reduce privacy risks, but it does not automatically resolve every ethical or regulatory question.

Researchers should distinguish genuinely anonymized information from pseudonymized or otherwise de-identified information that may remain linkable to individuals.

For example, a qualitative transcript may omit names but still identify an instructor through their unique institutional role, department, or involvement in a particular incident.

Even when data are effectively anonymized, institutional agreements or approved research procedures may restrict how the information can be processed.

The relevant assessment concerns the actual identifiability of the information, applicable permissions, and the proposed processing arrangement.

Researchers should not independently declare that anonymization eliminates all review obligations without considering the governing requirements.

Does Ethics Approval Cover AI Use Introduced Later?

An approved research protocol establishes authorization under specified procedures and conditions. It should not be interpreted as unlimited permission to introduce new processing methods.

Suppose a researcher initially plans manual qualitative analysis but later decides to use a cloud-based generative AI service to analyze complete participant transcripts.

The change may affect data recipients, storage arrangements, confidentiality, and the scope of participant information processed.

Depending on institutional requirements, the researcher may need to submit an amendment, request guidance, or obtain additional authorization before implementing the change.

The appropriate procedure depends on the significance of the modification and the original approval conditions.

Researchers should examine AI use introduced after research ethics approval before assuming the existing authorization is sufficient.

What About Research Conducted in the Philippines?

Researchers working in the Philippines should examine the requirements of their institution, applicable research ethics committees, and relevant national laws or guidelines.

The Data Privacy Act of 2012, Republic Act No. 10173, provides a legal framework for the processing of personal information. Its requirements may be relevant when research data are submitted to external AI services.

For health-related research, the Philippine Health Research Ethics Board and the National Ethical Guidelines for Research Involving Human Participants provide important ethical guidance.

However, requirements for educational, social, behavioral, and institutional research may also depend on the university's own research ethics policies and the nature of the study.

Researchers should not assume that the absence of a health-related intervention means there are no applicable ethics or data-protection requirements.

Nor should they assume that all research involving AI automatically requires the same review pathway.

How Should You Prompt AI to Help Assess Possible Approval Requirements?

AI can help organize the relevant questions before the researcher consults the appropriate institutional authority. It should not be asked to issue a definitive ethics clearance.

This prompt can help researchers prepare a more specific inquiry rather than asking an ethics committee whether "using ChatGPT" is generally allowed.

The institutional response, not the AI-generated assessment, should determine the applicable approval procedure.

Should AI Use Be Disclosed in the Ethics Application?

When AI is part of the proposed participant-facing procedures or the handling and analysis of participant information, researchers should consider whether it needs to be described in the application.

Relevant information may include the system's role, data to be processed, provider arrangements, safeguards, human oversight, and possible effects on participants.

Some institutions may require broader disclosure of AI assistance, while others may focus on activities that materially affect research procedures or participant risks.

Researchers should follow the applicable application instructions and avoid describing AI processing so vaguely that reviewers cannot evaluate its implications.

Detailed considerations concerning AI disclosure in research ethics applications should be distinguished from publication disclosure requirements, which serve a different purpose.

When Is the Best Decision to Ask the Ethics Office?

Researchers should seek an institutional determination when the proposed AI use involves uncertainty about participant consent, identifiable information, new data recipients, or changes to approved procedures.

For example, if an AI provider's terms do not clearly establish whether participant data may be retained or accessed, the researcher should not infer that the arrangement is acceptable merely because the service is widely used.

Similarly, if a study introduces an AI chatbot that provides individualized responses to participants, the researcher should clarify how the system fits within the approved intervention or data collection procedure.

Consulting the appropriate office before implementing a consequential change is generally preferable to attempting to justify an unauthorized activity afterward.

The purpose of consultation is to establish the applicable requirements, not to obtain a universal declaration that AI is either ethical or unethical.

Comparing Two Uses of ChatGPT in the Same Research Study

Both activities involve generative AI, but they raise different ethical questions.

The distinction demonstrates why ethics approval requirements should be evaluated according to the proposed research activity rather than the presence of an AI tool alone.

Common Misconceptions About AI and Research Ethics Approval

Ask Whether AI Changes the Research Procedures, Not Whether AI Is Present

Before using generative AI, describe the proposed activity precisely. Identify whether the system will process nonconfidential materials, participant information, or data requiring special protection.

Then examine the approved or proposed protocol, institutional requirements, and provider arrangements.

Document the AI system, proposed function, data involved, and institutional guidance obtained.

The objective is not to submit unnecessary ethics applications for every minor AI-assisted task. It is to ensure that consequential uses are appropriately evaluated and authorized before they occur.

Before Deciding Whether AI Use Requires Ethics Approval

Frequently Asked Questions About Generative AI Research Ethics Approval

Ethics Approval Depends on the Research Activity, Not the AI Label

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.