AI, Ethics, and Transparency in Research – Part 3
As generative AI becomes increasingly integrated into qualitative research, transparency in its use is important to maintaining ethical integrity and research credibility. In this final instalment of our trilogy on ethics in AI-augmented qualitative research, we discuss best practices for documenting AI’s contributions in research reports. This article provides practical guidance on what requires citation, what does not, and how to navigate the evolving standards of responsible AI use in scholarly work. By fostering transparency, researchers can enhance trust, uphold academic and professional standards, and contribute to the responsible evolution of generative AI in qualitative research.
The article begins by exploring various ways generative AI can support the research process, including literature review, data collection, data analysis, and writing. It then describes two approaches to ensuring transparency in AI use: citation and acknowledgment. This is followed by recommendations and guidelines from different publishers and universities, along with concrete examples of how to properly cite AI tools in academic work.
Literature Review
Systematic Review or Meta-Analysis
When conducting a systematic literature review—also known as a research synthesis or meta-analysis—the methodological approach to searching for and selecting sources is critical. In this context, the selected sources constitute the primary data material. Therefore, if an AI tool is used for sourcing or screening literature, its role must be carefully considered and documented as part of the methodological process. Following established citation guidelines, this means that AI tools should be cited in the same way as databases or any software used for analyzing the content of the selected sources.
General Literature Review
In contrast, when a literature review is conducted as part of an empirical research project—such as identifying research gaps or justifying the study’s relevance—it serves as a supporting rather than primary component of the research. In such cases, AI tools function similarly to reference management or search tools, assisting in the process but not directly shaping the research findings. As a result, citing AI tools is not required unless their use significantly influences the review’s conclusions.
Data Collection and Analysis
When using an AI tool to support data collection and analysis, it should be cited, as the choice of tools can significantly influence the content and outcomes of your research.
For data collection, document and cite the AI tool just as you would a research instrument or software. In research reports, specify the prompts used—this is comparable to detailing an interview guide—and discuss any limitations associated with using an AI bot for interviewing. Transparency in these aspects ensures the reproducibility and credibility of your methodology.
For data analysis, describe the analytical process in detail. If using conversational AI, save all chat transcripts, as they may be required by publishers or institutions when submitting your research. Even if you do not need to submit the prompt or the full chat, retaining full transcripts is crucial for reviewing the analytical process and ensuring transparency. These records will also be invaluable when writing up findings, as they allow for the inclusion of direct examples from the AI interactions.
AI support for writing
To uphold scientific writing standards, Hryciv et al. (2023) suggest the following approach for reporting AI involvement in manuscript development:
1. Specify the level of AI assistance – Clearly indicate how AI was used, whether for proofreading, language refinement, or more advanced tasks like content generation.
2. Clarify human involvement and oversight – Describe the extent of human contribution throughout the writing process, including editing, revising, and verifying AI-generated content.
3. Provide details about the AI tool – Mention the AI tool or model used, including its name, version (if applicable), and primary functions.
4. Address ethical considerations – Disclose any ethical implications, such as potential biases introduced by AI, and how they were managed.
5. Acknowledge limitations – Recognize any shortcomings of AI-generated content, including possible biases, inaccuracies, or misinterpretations.
To better categorize AI’s role in manuscript development, Hryciv propose the following classification system for describing the extent of AI assistance in the writing process:
- Proofreading
- Restructuring
- Drafting
Proofreading
Proofreading with AI means that the human author is responsible for the content, while an AI tool is used to ensure grammatical accuracy and correct syntactic errors. AI-assisted proofreading helps refine sentence structure, fix minor mistakes, and improve clarity without altering the author's intended meaning.
Example 1 (Grammar)
Version 1 (Human author): "The researcher analyse the data and found interesting pattern."
Version 2 (AI-corrected): "The researcher analysed the data and found an interesting pattern."
Example 2 (Clarity and readability improvement)
Version 1 (Human author): "Due to the fact that the experiment was conducted in a controlled environment, it was possible for the researchers to obtain reliable results."
Version 2 (AI-corrected): "Because the experiment was conducted in a controlled environment, the researchers obtained reliable results."
These examples demonstrate how AI can enhance grammar, clarity, and readability while maintaining the original intent of the author.
Restructuring
Text restructuring with AI goes beyond basic proofreading. In this process, AI is used to improve or reformulate sentences while preserving the original meaning. The researcher remains responsible for content creation, ensuring that ideas and arguments are their own. After AI-assisted revision, the text is integrated back into the author’s work and further refined to align with their style and intent. The original version remains recognizable in terms of content, but the phrasing is enhanced for clarity, readability, or conciseness.
Original text (researcher)
"Yesterday I decided to take a long walk in the park. The sun was shining, the birds were chirping, and it was the perfect day to go outside. During my walk, I thought about different projects I was working on and how I could improve them. I also saw many families and friends enjoying the beautiful weather. It was really a refreshing day."
Restructured text (AI proposal)
"Yesterday, I set out for a long walk in the park, drawn by the warm sunshine and the cheerful chorus of birds. The day felt perfect for being outdoors. As I strolled through the vibrant scenery, my mind wandered to the various projects I’m working on, sparking new ideas for improvement. Along the way, I noticed families and friends making the most of the beautiful weather, adding to the lively atmosphere. The walk left me feeling refreshed and inspired.”
Finale refined version
LLMs can sometimes be a bit too creative with words and at times, it is a bit over the top (depending on taste of course). Thus, you may want to adapt it slightly. The final version could look like this:
"Yesterday, I set out for a long walk in the park, drawn by the warm sunshine and the cheerful singing of the birds. The day felt perfect for being outdoors. I was not alone, many others like groups of friends and families were also making the most of the beautiful weather, which created a lively atmosphere. With all of this in the background, my mind wandered off to the various projects I’m working on, sparking new ideas for improvement. Being back home, I felt refreshed and inspired.”
Drafting
Here, the AI generates a working draft based on input data and human-written instructions. This can also go beyond writing, e.g. for drafting an interview guide. The human author then significantly revises the AI-generated content, creating a clear difference from the initial AI output.
Task given to an AI-Assistant by a researcher
I want to investigate the effects of working from home on job satisfaction. My research questions are:
- How has the switch to working from home affected your daily work routine?
- What advantages and disadvantages do you see in working from home?
- To what extent has working from home influenced your job satisfaction?
Develop an interview guide that captures the employee perspectives.
AI-generated suggestion
Introduction: Start the interview with a brief introduction and explain the purpose of the study. Emphasize the importance of respondents' honest and open expressions of opinion.
Body: Work routine: "Tell me about your daily work routine in the home office. How does this compare to your routine in the office?"
- Pros and Cons: "Can you name specific pros and cons of working from home?"
- Job satisfaction: "How has your perception of your job satisfaction changed since you switched to working from home?"
Closing: Ask participants if they have any other comments or would like to add something that hasn't been addressed before. Thank them for their time and contribution to the study.
Interview Guide put together by the researcher based on AI Input:
Making the Use of Generative AI Transparent
Citing or Acknowledging
Citing refers to formally referencing an AI tool when it has directly contributed to the research process or findings. This is similar to citing a software program, dataset, or published work that influenced your analysis.
Acknowledging an AI tool is appropriate when the tool provided support but did not directly contribute to the main body of the research findings. For example, this applies when an AI tool is used to assist with finding and analysing literature for the literature review section of an article or thesis. An acknowledgment at the end of an article can look like this:
“The authors of this manuscript declare that in the writing process of this work, no generative artificial intelligence (AI) or AI-assisted technologies were used to generate content, ideas, or theories. We utilised AI solely for the purpose of enhancing readability and refining language. This use was under strict human oversight and control. After the application of AI technologies, the authors carefully reviewed and edited the manuscript to ensure its accuracy and coherence” (Khalifa, and Albadawy, 2024).
As a general rule of thumb, Kristing (2024) suggests: If AI tools have a direct impact on the methodology or results, cite them. If they merely support the process, no citation is required:
Guidelines by Publishers and Universities
An increasing number of publishers and universities are issuing guidelines. Therefore, it’s a good idea to check whether your university or the publisher you plan to submit to has already released any. Sage Publications for example makes a distinction between AI assisted tools, and AI generated content. They write:
“We believe that AI-assisted writing will become more common as AI tools are increasingly embedded within tools such as Microsoft Word and Google Docs. You are not required to disclose the use of assistive AI tools in your submission, but all content, including AI-assisted content, must undergo rigorous human review prior to submission. This is to ensure the content aligns with our standards for quality and authenticity.” Source
The American Psychological Associations, however, wants you to cite the AI model when you used one for drafting a manuscript:
“When a generative AI model is used in the drafting of a manuscript for an APA publication, the use of AI must be disclosed in the methods section and cited.” Source
Thus, before making the decision to publish in a particular journal, it is advisable to inform yourself in detail about any specific guidelines and requirements that may exist in this journal.
Here are some links to the AI policies of various publishers and universities:
Sage Publications AI Guidelines
For a comprehensive list of AI Publisher Policies, see Purdue University’s AI page.
Guidelines in German
Berlin Universities Publishing
Universität Basel: Leitfaden «Aus KI zitieren»
Universität Hamburg: Leitfaden zum Umgang mit KI-Tools in Bachelor-, Master- und Seminararbeiten
Universität Wuppertal: Wie belege ich die Verwendung von KI-generierten Inhalten?
In comparing the above listed guidelines, I’ve noticed that many draw from or reference those that were published first, such as APA policies. It seems that a common standard is emerging, and in the near future, my best guess is that thes guidelines will likely become more uniform across publishers.
How to Cite
Citing ChatGPT / Claude (and other chatbots)
Format (APA Style): Author. (Date). Name of tool (Version of tool) [Large language model]. URL
Examples:
OpenAI (2025). ChatGPT (May 14 version) [GPT-4o]. https://chat.openai.com/chat
Anthropic (2025). Claude (November 22, 2024) [3.5 Sonnet]. https://www.anthropic.com/
In-Text Citation Example:
(OpenAI, 2025)
(Anthropic, 2025)
Citing Content Generated by Artificial Intelligence
Format (Chicago style): Author, Title, Publisher, Date, url for the tool.
Example (if information about the prompt has been included within the text of your paper):
Text generated by ChatGPT, OpenAI, May 14, 2024, https://chat.openai.com/chat.
If the prompt has not been included in the text – as note:
ChatGPT, response to "Explain how to make pizza dough from common household ingredients," OpenAI, May 14, 2024, https://chat.openai.com/chat.
Citing Tools for Data Collection and Analysis
Below are some examples. You can follow this format and replace the names, years, and links with the relevant details for the tools you are using.
Causal Map Ltd (2025). QualiaInterviews. Retrieved XXX from https://www.qualiainterviews.com.
VERBI Software (2025) AI-Assist in MAXQDA 24. Berlin: Verbi GmbH. https://www.maxqda.com.
QInsights BV (2025). QInsights [software]. The Netherlands. https://www.qinsights.ai
In-Text Citation Example:
We used QualiaInterviews (Causal Map Ltd, 2025).
We used MAXQDA AI Assist …. (MAXQDA 2024, VERBI software)
We used QInsights for data analysis…. (QInsights BV, 2025)
Resources & References
Gregor, S. (2024). Responsible artificial intelligence and journal publishing. Journal of the Association for Information Systems, 25(1), 48-60. https://doi.org/10.17705/1jais.00863
Hryciv, B. N., Seely, A. J. E., & Kyeremanteng, K. (2023). Guiding principles and proposed classification system for the responsible adoption of artificial intelligence in scientific writing in medicine. Frontiers in Artificial Intelligence. https://www.frontiersin.org/articles/10.3389/frai.2023.1283353
Khalifa, M., & Albadawy, M. (2024). Using artificial intelligence in academic writing and research: An essential productivity tool. Computer Methods and Programs in Biomedicine Update, 5. https://www.sciencedirect.com/science/article/pii/S2666990024000120
Terill, K. (2024). Should I cite the AI tool that I used? Academic Insight Lab Blog. https://academicinsightlab.org/blog/how-should-ai-tools-be-cited-or-acknowledged
Academic Insight Lab. (2024, January 10). How to acknowledge generative AI in academic writing [Video]. YouTube. https://www.youtube.com/watch?v=2eKkGuRFR80