Instructional Designers’ Integration of Generative Artificial Intelligence into Their Professional Practice

Abstract

Integrating generative artificial intelligence (GenAI) into professional practice has become an important topic for professional instructional design practice and training. Accordingly, the purpose of this multiple-case study was to examine six professional instructional designers’ integration of GenAI into their professional practice and the factors affecting this integration. Research data were collected through semi-structured interviews conducted with professional instructional designers working in corporate or higher education settings. The results were as follows: (a) instructional designers mostly integrate GenAI into instructional design and/or development phases and they think that it also has the largest impact on these two phases; and (b) instructional designers’ integration of GenAI into their professional practice is mainly based on their ambivalent attitudes toward it, which is closely linked to the advantages and disadvantages associated with the technology. Specifically, instructional designers’ basic understanding of GenAI, the efficiency of generating instructional content through GenAI, the inaccuracy of GenAI-created products, instructional designers’ use of GenAI in everyday life, and institutional or company support shape their attitudes towards and integration of GenAI into their professional practice. All these findings suggest that instructional design and development phases are especially vulnerable to and can benefit from instructional designers’ attitudes and use of GenAI. Accordingly, it can be useful to address and enhance attitudes toward GenAI technology in instructional design training, which can promote instructional designers’ acceptance of the technology and effective use of it.

1. Introduction

The emergence of ChatGPT (GPT-3.5) in November 2022 immediately drew attention to the use of artificial intelligence and large language models, and how they can be used in various fields including instructional design (e.g., Kumar et al., 2024). Integrating artificial intelligence (AI) into instructional design (ID) and using it to automate ID practices has been researched in the past. R. Tennyson (1984) described using AI in an adaptive instructional system to provide real-time personalized instruction, and Winn (1987) also explored ways of making instruction more adaptable with AI. Additionally, R. D. Tennyson and Barron (1995) covered intelligent systems to automate planning, production, and implementation, and addressed implications, added value, and application issues. During the 1990s, several automated ID systems were developed, such as advisory systems, expert systems, information management systems, performance support systems, and authoring tools (Kasowitz, 1998).

Large language models (LLMs) can facilitate various activities, such as developing learning materials, information seeking, and synthesizing or summarizing large amounts of information, and can therefore be useful in education and training. LLMs have seen widespread use in educational settings recently (e.g., Chassignol et al., 2018; Egara & Mosimege, 2024), even though there is a strong precedent of AI integration into education prior to GenAI (e.g., Hwang et al., 2020; Popenici & Kerr, 2017). Essentially, AI can let an instructional designer (IDer) focus on more complex and creative tasks by handling more mundane and time-consuming labor (Escribano et al., 2024; Ch’ng, 2023; McNeill, 2024). In this sense, AI can also be helpful for others involved in designing instruction. For instance, by automating routine administrative tasks, teachers can spare more time for better ID by developing more interactive and engaging learning content (Deshpande et al., 2023).

Despite the emerging findings and inferential insights in relation to IDers’ integration of GenAI into their professional practice, we still need to know more about how this integration actually happens, and the lesser-known factors that affect this integration. Gaining such insights can help us articulate the status quo and make evidence-informed improvements including innovative AI-powered ID tools, models, or work processes that would enhance not only efficiency but also effectiveness. This knowledge can also help us make informed decisions regarding integrating AI into IDers’ training, thereby better preparing the future ID workforce. Accordingly, this exploratory multiple-case study intends to uncover the participating six IDers’ integration of GenAI into their professional practice and the main factors affecting this integration.

Instructional Design and GenAI

This study is conceptually informed by instructional design as a systematic process of designing, developing, and evaluating learning experiences to enhance learning (Dick et al., 2022; Smith & Ragan, 2005). ID focuses on creating instructional solutions to which GenAI can contribute (Parsons & Curry, 2023) by, for example, ameliorating tedious tasks and optimizing more creative tasks (Bolick & da Silva, 2024; Ch’ng, 2023). To illustrate, establishing learning objectives and then creating a video that aligns with those objectives can be considered a routine and perhaps tedious task given the amount of time and effort needed. However, relating those learning objectives and the video to the next curriculum topic might be considered more creative since the flow of instruction can go in different directions resulting in different learning outcomes. Thus, GenAI can take over content creation while IDers can focus more on quality (Bolick & da Silva, 2024). GenAI can also enrich ID in other ways by enhancing students’ understanding of ethics, limitations, and challenges (Wood & Moss, 2024), and performing needs, task, and learner analyses despite potential limitations (Parsons & Curry, 2023). Parsons and Curry’s (2023) results indicated that ChatGPT is successful at completing ID tasks based on the data it was pre-trained on but did not address more specific details regarding learning context and learner profiles.

Analysis, design, development, implementation, and evaluation (ADDIE) is a generic and rooted ID model, thereby functioning as a reference framework (e.g., Ch’ng, 2023). According to Ch’ng (2023), GenAI can achieve the following: (a) enhance analysis by expediting data collection and analysis; (b) help with design and content creation; (c) promote self-directed learning; and (d) enhance authentic assessment and evaluation. In this regard, Escribano et al. (2024) proposed a GenAI tool that would automate big data analysis thus producing findings and suggestions for IDers to review and revise. Even before ChatGPT, Hwang et al. (2020) stated that AI can help design better learning activities and develop better technology-enhanced learning. GenAI visual creation tools can also produce content thus promoting design ideation (e.g., Huang et al., 2024). ChatGPT can enhance the media selection process for IDers and educators (DaCosta & Kinsell, 2024). Szilas and Emery (2024) used GenAI to turn videos into text creating textual interactive learning content. Given such examples, Ch’ng (2023) claimed that GenAI can make the ADDIE processes more effective and efficient.

IDers’ collaboration with GenAI can be helpful for instructors, students, and the IDers themselves. Comparing instructional materials created by GenAI and humans, Oteng-Darko et al. (2024) concluded that GenAI can help with ID to create better resources, but AI work should accompany human oversight. Likewise, Katiyar et al. (2024) suggested a collaboration among IDers, educators, and GenAI developers to align personalized learning with curriculum rooted in ID and learning theories. However, Krushinskaia et al. (2024) warned that collaboration with GenAI may be problematic since GenAI would try to please users and change its answers depending on whether users would like them or not.

GenAI also seems to be promising in education in relation to ID. In health education, for instance, GenAI can help create content in line with curriculum and enhance learning (Houssani et al., 2024) and promote personalized learning and higher-level skills (Sallam, 2023). Likewise, GenAI-created content can be used in design education to enhance student self-efficacy and higher-order thinking during design ideation (Huang et al., 2024). In music education, GenAI can support educators with curriculum alignment and effective ID (Quian, 2023). In general, GenAI can enhance personalized or adaptive learning by tailoring content and instruction based on learners’ needs, preferences, and/or capabilities (e.g., Chen et al., 2020; Gligorea et al., 2023; Quian, 2023; Willis, 2023) as well as assessment and intelligent tutoring systems (e.g., Bahroun et al., 2023). Omar et al. (2024) called it personalized ID, in which GenAI can enhance professional development and training.

Since 2022, GenAI tools such as ChatGPT have been publicly available, affording spontaneously generated images or text from simple prompts (Bolick & da Silva, 2024). Applying such tools for content creation in ID is obvious, but GenAI can also perform other sophisticated tasks including brainstorming ideas (Barrett & Pack, 2023), reasoning (Kojima et al., 2023), and creating assessments (Aboalela, 2023). However, LLMs that drive GenAI are non-deterministic (S. Ouyang et al., 2023), susceptible to algorithmic bias and hallucinations (Buolamwini & Gebru, 2018; J. Luo et al., 2024), and have demonstrated fluctuations in reliability over time (Pack et al., 2024). Still, early adopters are exploring GenAI use in ID.

B. Nguyen et al. (2023) demonstrated GenAI-assisted content creation for developing training materials, and found that using GenAI to generate text, images, and audio for training materials can lead to considerable reductions in time and money expenditures; however, monitoring and revising is necessary. Authors also speculated that as GenAI becomes more efficient at content creation, traditional ID skills surrounding content creation such as graphic design, information synthesis, and summarization, may be eclipsed by skills such as monitoring, checking, and editing. Likewise, B. Nguyen et al. (2024) suggested that resources such as time and effort may need to be shifted to front-end analysis, or to novel tasks such as prompt engineering and GenAI tool selection. Additionally, Binhammad et al. (2024) highlighted the utility of GenAI for creating personalized learning materials, which is typically prohibitive due to the cost and time required.

GenAI has also been employed in brainstorming, outlining course structure and learning objectives, creating assessments, and analyzing learner progress (McNeill, 2024). Kumar et al. (2024) and T. Luo et al. (2025) both remarked that brainstorming is a frequent use of GenAI by IDers. Using GenAI to draft learning objectives or generate ideas for learning activities can help IDers overcome designer blocks (e.g., T. Luo et al., 2025). Kumar et al. (2024) also highlighted using GenAI to augment subject matter experts’ (SME: experts having content area or subject area know-how) input. Additionally, IDers use GenAI to create assessment rubrics (Kumar et al., 2024), and others have shown that GenAI can reliably assess learners’ writing using a rubric (Pack et al., 2024). Rezigalla (2024) used ChatPdf to generate multiple choice questions (MCQs) and then had them rated by SMEs, with results showing that SMEs rated the items as excellent or good. However, Ngo et al. (2024) used GPT-3.5 to generate MCQs and reported the need for substantial revisions due to incorrect answers. GenAI has also proven useful in learning analytics, with various machine learning techniques providing instant and adaptive feedback on learner performance (F. Ouyang & Zhang, 2024). Escalante et al. (2023) provided both instructor and GenAI-created feedback to learners and reported that learners valued both equally.

IDers engage in multiple professional tasks, and research and scholarly work presenting opinions suggest that GenAI can be usefully integrated into them. However, given the perceived strengths, opportunities, weaknesses, and threats (e.g., cheating) associated with GenAI tools such as ChatGPT (Na et al., 2024), and that the technology is still new and has been developing very rapidly, IDers may already be employing the technology in newer and unexpected ways in their professional practice. More importantly, less is known about the factors that affect the integration of GenAI into professional ID practice by IDers. There is a need, therefore, to better understand not only how GenAI is integrated into professional ID practice but also the corresponding factors that affect this integration, and how this integration works thereby adding to the previous findings. To this end, this study addressed the following questions:

• How have instructional designers been integrating GenAI into their professional practice since the launch of ChatGPT?
• What factors have been affecting instructional designers’ integration of GenAI into their professional practice?

2. Methods

2.1. Research Design

This is an exploratory multiple-case study (Yin, 2014, 2018), which provides a purposeful analysis of two or more single cases (Stake, 1995) as well as a cross-case analysis (Stake, 2006). A multiple-case approach was chosen since it can help us understand a phenomenon from multiple perspectives without losing insight into each case’s unique aspects (Thomas, 2011). The researchers stopped at the sixth case when they decided that data saturation occurred or when no new codes and no themes emerged from the research data, following Guest et al. (2006). Among the case study data collection methods provided by Yin (2018), this study employed semi-structured interviews. Lastly, a case refers to an individual IDer, and each one was chosen based on the following criteria: (a) having a current active or passive position; (b) years of ID experience; (c) years of experience related to ID; and (d) highest educational degree. Each case was assigned a pseudonym for confidentiality.

2.2. Participants

Participants were four female (67%) and two male (33%) IDers with an age range of 31–56 (M = 44.75; SD = 9.35) (

Table 1. Participant profile.

Pseudonym	Position	Years of ID Experience	Years of Experience Related to ID	Highest Degree
Nikki (Case 1)	IDer in a corporate	6	>20 years newspaper editor (online)	BA—journalism + grad certificate—ID
Taylor (Case 2)	Senior IDer in a corporate	20	5 years—technical writing	training and performance improvement
Elvin (Case 3)	IDer in higher education	5	5 years— K-12 teacher	PhD—learning, design, and technology
Cameron (Case 4)	Postdoc researcher	12 (higher education + contract-based)	4 years—outreach coordinator	Postdoc—learning, design, and technology
Jordan (Case 5)	IDer in a corporate	3	5 years—onboarding new staff	EdD—learning design and performance technology
Riley (Case 6)	IDer in higher education	10	none	PhD—instructional systems technology

).

2.3. Instruments

2.3.1. Demographic Survey

The participants completed a demographic survey (Appendix A) delivered via Qualtrics thus providing general descriptive participant profile insights.

2.3.2. Semi-Structured Interviews

We employed individual semi-structured interviews to understand participants’ integration of GenAI into their professional practice and the factors affecting this integration. The questions were originally prepared based on projected ID experience and whether participants would be integrating or planning to use GenAI (Appendix B). Based on a pilot interview, some interview questions were revised to use ADDIE language to better understand how participants expected GenAI to affect ID.

2.4. Procedures

2.4.1. Data Collection

After Institutional Review Board (IRB) approval, IDers were invited to contribute via social media and an IDers’ group email list in a southeastern university in the U.S. The following step was a pilot interview with an IDer to check how the interview protocol would work. Since the main inclusion criteria were having 5 or less than 5 years, or more than 5 years of professional experience, those who wanted to participate in the study were contacted again. Namely, participants who completed the demographic survey were contacted via email to set up interview times and dates. All interviews were performed on Zoom, recorded and transcribed verbatim automatically on Zoom, and saved anonymously. Each interview took approximately 40–60 min. Two researchers from the team cleaned interview transcripts.

2.4.2. Data Analysis

Two researchers independently coded the first interview data inductively to create initial codes and themes, and a third researcher compared their codes one by one and noted the number of agreements and disagreements to calculate the agreement percentage. There was 86% agreement, and the three researchers came together and discussed the codes and themes, achieving full consensus and establishing a coding scheme including initial codes consisting of reactions (positive and negative), organizational support, competencies, challenges, and ID phases (analysis, design, development, implementation, and evaluation). Specifically, the unit of analysis was derived from meaning, based on sentences, phrases, or paragraphs. Coding was performed using MS Word in a shared cloud-based folder. Overall, following Saldana (2016), the first coding included the following: (a) reading through the transcript and inductively coding to identify preliminary codes; (b) reading the transcript again and creating additional codes if necessary; and (c) finalizing the coding scheme.

The two researchers who coded the first interview data also coded the second interview using the coding scheme. They were also allowed to create any new codes as they seemed appropriate. The two coders then came together with the same third researcher to discuss the results. There was an initial agreement of 93%. The three researchers also agreed to eliminate three codes (i.e., reactions, challenges, competencies,) and add five new codes (i.e., attitudes, understanding of AI, efficiency or usefulness, inaccuracy, using AI in daily life) to the coding scheme. In the third round, the same two coders coded all four remaining interviews, and the same third researcher compared their coding. The lowest initial agreement level was 24%, while the highest disagreement level was 26%. Next, the three researchers came together and discussed all the codes and came up with the final ones. The final overall inter-coder agreement level was 93%, which indicates a good level (Creswell, 2014), and full consensus was achieved through discussion.

2.5. Trustworthiness

For confirmability, we pilot-tested the interview questions and revised them to increase clarity, comprehensiveness, and relevance thereby increasing their face and content validity. Likewise, some initial interview questions were also prepared based on existing ID know-how including ADDIE. To achieve a high level of dependability, the following was conducted: (a) two researchers coded transcripts; (b) a third researcher checked and compared their coding; (c) all three researchers came together to discuss codes and followed a coding scheme that was updated when needed through full consensus. Namely, multiple researchers were involved in the research procedures for consistency and researcher triangulation (Yin, 2018). This presentation of the thick descriptions of the research context and procedures provides transparency for replication purposes, thereby increasing transferability. We also selectively presented direct quotations from interviews to clarify participants’ main points.

As for reflexivity, the research team shares some background similarities with the participants. For instance, one researcher has a doctoral degree in learning, design, and technology, and other researchers are doctoral students in an instructional systems program with a similar ID training. Such similar backgrounds and experiences would increase credibility and help the researchers analyze and report the participants’ insights accurately (Yin, 2014). Finally, we think that informed GenAI integration into ID is important for practitioners and their educators to achieve enhanced learning.

3. Results

Results indicated the following: (a) the participating IDers have been integrating GenAI into their professional practice in various ways, but mostly for design or instructional planning and development purposes; and (b) IDers’ attitudes toward GenAI are the main factor that affects their AI integration into professional ID practice, which closely relates to other factors ranging from their understanding of the technology and their institutional or organizational support.

3.1. Instructional Designers’ Integration of GenAI into Their Professional Practice

3.1.1. Nikki

Nikki said her institution encourages using GenAI in ID. With granted access to ChatGPT, IDers are asked to work on research and development to use GenAI. Although her institution offers full support, proprietary information is heavily protected. Nikki also shared various ways of using GenAI in her ID work based on ADDIE. In analysis, she used the tool to conduct thematic analysis, inputting some data, such as user interviews or big institutional data, and generated themes, and highly appreciated the results.

In design, Nikki uses GenAI to solve design problems. She also said that GenAI is useful for visual, contextual, and textual design, and acknowledged that it is impactful. In addition, she deemed GenAI useful in mapping or alignment; as her institution used a competency-based education system, GenAI assists her in aligning objectives with competencies. In development, Nikki heavily uses GenAI for scenario generation. She also created a considerable number of assessments, learning activities, and supplemental resources. Additionally, she uses GenAI to develop ideas and generate contexts for assessments. For instance, she receives help with narratives and storylines when creating case studies. She also shared that GenAI can also be used for coding or programming.

In implementation, Nikki indicated that GenAI could have some impact although it still seems minimal. She thought implementation entails more human judgment and cultural or contextual considerations. She said, “…implementation may be the one that may be a little bit more human”. In evaluation, Nikki said GenAI can analyze data easily, thus linking evaluation and analysis, and acknowledging its usefulness. Nikki also said that she had been exploring ChatGPT as a student resource to support their learning.

Nevertheless, Nikki also indicated some limitations, claiming that ChatGPT did not perform well in mapping objectives. She even described it as “a little bit troubling, or at least problematic”. Thus, she indicated that human judgment plays a significant role in design. Additionally, she highlighted the need for editing and refining the content generated, and she highlighted educational background to be a meaningful supporter helping her evaluate technologies including GenAI: “…I think that my specific background in educational technology and instructional design has helped me to see these cycles of technology implementation, development that, maybe, other fields have been, you know, aren’t quite as aware of”.

3.1.2. Taylor

Taylor’s institution supports employees’ GenAI use and encourages IDers to learn about GenAI tools by testing them on the job. Still, the company restricts their employees from feeding proprietary information to GenAI. They also have a task force researching GenAI, providing corporate-level guidelines, and addressing any questions or concerns. Taylor also shared her colleagues’ GenAI use in everyday tasks, including writing emails.

Taylor is not heavily involved in analysis, implementation, and evaluation. Her major task is creating educational materials, and she finds ChatGPT helpful in writing textual resources. It also helps her brainstorm and develop content as it generates interesting phrases and ideas. For instance, Taylor used ChatGPT to create a periodical event title and its website, which helped her save time and work productively. She also anticipated that GenAI could be used to write quizzes or diagnostic questions to evaluate learners’ knowledge, offloading SMEs’ tedious tasks.

While benefiting from GenAI, Taylor always checks the quality and modifies GenAI-created ideas and content because she feels some texts can be repetitive. Another perceived limitation is grounded in some common tendencies and learners’ preference: “…people are still very attached to in-person learning and they really want, you know, they wanna be able to interact with the person giving the training”.

3.1.3. Elvin

Elvin’s institution also encourages GenAI use. The university developed its own GenAI tool, in the testing stage, and the goal is to utilize it to save time and resources in ID. His institution is also expecting big future opportunities that GenAI will bring. Elvin also has concrete insights into using GenAI for professional purposes and using ChatGPT in ID but not in every ADDIE phase. In analysis, Elvin said it can be used for taking, organizing, and analyzing notes; making a voice memo and having GenAI transcribe it into notes, and these notes could be used for data analysis purposes.

In development, Elvin uses ChatGPT to generate groundwork to start writing an article. Elvin shared the instance when GenAI helped overcome writer’s block as it assisted with giving him a kick start. After paraphrasing and editing the writing, Elvin successfully completed it. Elvin also said GenAI can be used for administrative tasks in ID. For example, he said it could be useful in implementation because the tool could reduce manual efforts that are needed to upload course modules to learning management systems.

Elvin also highlighted some limitations by pointing to the need to review and edit, and emphasized that human creativity could not be replaceable by GenAI: “It’s not going to add that extra nugget that a human creating the same content would add”. Elvin still thought GenAI could perform creative work, but it is task oriented. However, the human element can spice up creativity, which is more than completing ID tasks.

3.1.4. Cameron

Cameron and her colleagues are asked to use GenAI in ID by their institution, but they are not allowed to feed proprietary information into it. The institution also has internal SMEs, someone external to review the generated content, and provides guidelines and recommendations for specific prompts.

Cameron does not think GenAI is a good tool to conduct data analysis. Hence, Cameron uses it to achieve simpler analysis tasks, such as generating data summaries. However, Cameron thinks GenAI is useful for course development because they do not need to start from scratch; Cameron uses GenAI to generate ideas, content information, scenarios, questions and answers, videos, voice files, images, and summaries of materials. Cameron highly values GenAI’s contribution in idea generation since it helps her get started: “It goes beyond brainstorming. It’s really like storyboarding in a way, scripting, expanding. So, you really kind of get that initial skeleton…”.

Cameron also said that GenAI is good at creating interactive scenarios and questions and appreciated that it saves time for thinking about how to word questions. Cameron also uses GenAI to select the most appropriate videos for a given script. Although the video product may not be perfect, Cameron easily edits and updates it. Likewise, Cameron generates voices using ElevenLabs and edits it, and finds it useful because it can generate audio, and GenAI-generated voices sound like a human’s. Instead of hiring or asking people to record narrations, GenAI creates what Cameron needs within a few minutes. She is also planning to develop chatbots using GenAI.

Cameron also uses GenAI to enhance the existing materials and course content. She said, “Another thing that I like about it, sometimes faculty would have some objectives already. And I would take theirs, and also take the content, and see if I can improve…”. Similarly, Cameron explores GenAI to write better and increase accessibility.

Cameron thinks using GenAI can have limitations in implementation. In evaluation, Cameron uses GenAI to create assessment items and rubrics, which is also related to development, and benefits from brainstorming ideas for evaluation assignments. In addition, she shared that she explores the tool Pando, which analyzes behavior data and suggests how it could change. Since it helps with how to adapt learning strategies and vision strategies based on the behavior data, Cameron showed curiosity and anticipation to use it for evaluation purposes.

Cameron highlighted the cost and time efficiency associated with GenAI in ID despite some limitations; GenAI provides general responses lacking details and clarifications. Therefore, Cameron thinks prompting skills are important. Cameron also needs to review the quality of responses in respect to accuracy and reliability: “You really have to be careful, because it’s a language generation model, you know, it can create some nice sound and text without it being reliable”. Accordingly, Cameron double-checks and compares GenAI-created items with the original sources and continuously modifies prompts to improve the output.

Cameron also needs to edit content created by GenAI and thinks GenAI’s role is drafting content. It is the IDers’ job to edit and develop that content further: “It’s a good start, but it’s not always perfect”. Similarly, regarding visuals, Cameron prefers to create basic flow charts since the output may not always be successful. Despite these limitations, Cameron thinks using it with caution and employing personal judgments can be safe. Cameron added that knowledge and skills in research and literature review help evaluate the accuracy and the quality of GenAI-created content. Nevertheless, Cameron anticipates that GenAI’s limitations will be improved.

3.1.5. Jordan

Jordan’s institution does not use GenAI and has strict intellectual property policies employees are required to follow. Although institutional concerns about GenAI use and copyright infringement were not formally communicated, Jordan is aware of possible negative impacts: “Because you do want to protect your craft in a way. If everybody had all the keys to the kingdom, there would be no competitive market, there would be no, you know, nothing that sets you apart from other organizations”.

Jordan still uses some software including beta GenAI such as Adobe programs in his tasks. In design, Jordan uses ChatGPT to brainstorm ID strategies. In addition to ID textbooks as a job aid, Jordan uses GenAI to review and confirm ID solutions. After a training is delivered, Jordan would ask GenAI to create ideas for alternative instructional strategies based on the performance metrics, approaches used, information about implementation, and the outcomes. Jordan treats GenAI as an assistant helping him come up with ideas to solve ID problems and improve instruction and thinks GenAI will impact design most, but foundational ID knowledge necessary for development is relatively unchanged: “I think design. Specifically, when it comes to different approaches and models that can be used to develop the training, I don’t necessarily think [they] will be impacted by AI, simply because there are tools that we use to develop instruction, and those have remained unchanged for the past decade”.

In development, Jordan reported using GenAI to build lessons and relevant instructional content, generating training scenarios and looking for suggestions for developing instruction. In evaluation, Jordan shared that GenAI would impact “redesigning and updating based on the results of initial evaluations and then redeploying”. Jordan explained that identifying assessment levels and measurable performance metrics can be completed by human upfront and GenAI could help revamp instructions.

Jordan also shared accuracy, ethics, and reliability concerns about GenAI since it can offer various information and resources without fact-checking. Thus, Jordan thinks that effective prompting and critical evaluation of the output is essential. To this end, Jordan stressed the importance of ID foundational knowledge: “…they still have to know how to complete a task analysis in order to be able to tell the GenAI what they’re looking for in order to specify, you know, draft a scenario for this lesson”. Thus, Jordan also thinks that using GenAI should not mean slacking because IDers should conduct a goal analysis and evaluate the alignment between goals and ID solutions before confirming them with GenAI. As such, Jordan values IDers’ contribution and judgment, which is irreplaceable: “You (AI) don’t have that core component of understanding models, approaches, strategies. The more in-depth learner analysis. You don’t have that feedback loop through evaluation and restructuring, redesigning”.

3.1.6. Riley

Riley’s institution encourages GenAI use in ID and asks IDers to be seasoned in using GenAI efficiently and effectively to create better learning experiences. Hence, the institution provides various professional development opportunities, such as webinars and demonstrations by GenAI companies. Additionally, they have regular meetings where the instructional technology department offers GenAI-related presentations and demonstrations to enhance work efficiency and quality.

Riley shared various examples for adopting multiple GenAI tools (e.g., ChatGPT, Gemini, Microsoft Copilot) in daily life and job tasks. Riley also uses Reddit and Google to search for information, but says that ChatGPT retrieves information and resources more quickly. So, Riley thinks GenAI is helpful in obtaining quick answers. In ID, Riley’s GenAI use includes design and development. Riley also added that GenAI produces responses only based on user input, which results in broad answers: “AI’s insights are usually generic, so it cannot really know about learners. Task analysis and such, so, you can still use AI for those too, but it won’t have the clear picture of your specific audience”.

In design, GenAI assists Riley while brainstorming for learning objectives and images. GenAI generates useful suggestions for learning objectives once Riley provides the teaching and learning context and instructional goals. Even though the output is not perfect, it saves time for Riley to help the faculty in writing better learning objectives. Riley also reported that GenAI is useful for generating image ideas. Riley uses ChatGPT or Gemini to receive suggestions regarding what images would be appropriate. Then, Riley picks the suitable ones and finds them on stock image websites. Again, GenAI may not always give the best options; however, it helps Riley be more creative.

Riley is also satisfied with using GenAI to create an outline and generate ideas for lesson planning and course activities. She said, “AI does a much better job than I would do as an instructional designer with coming up with ideas”. She also thinks that it goes beyond brainstorming: “… I can use the help of AI, GenAI. Again, I edit them. So, it’s different than brainstorming because I tell it what to do and I use a version of what they create after editing. So, it’s very helpful in that sense too.”

In development, Riley noted that GenAI can develop an instructional outline, assessments, alt-texts, and activities: “We have to provide alt text for the content because of the accessibility requirements. So, I take a screenshot of the flow chart or whatever the instructor has and enter it to Gemini. Gemini is better in that”. Moreover, Riley creates course activities including ice breaker prompts and activities that are relevant through GenAI. Additionally, Riley uses GenAI for editing recorded narrations: “We also have Descript, a tool that we can use to generate voiceover, and I also edit some content to some extent. (…) To copy the voice role of the faculty to, you know, to fix things in the recording. You can use Descript to mimic the voice of the instructor so that you don’t have to have the faculty to really record it”.

Creating summaries of faculty meeting notes also works well for Riley to save time and Riley’s team uses GenAI to make faculty’s slides more engaging, creative, and visually attractive. Riley also thinks GenAI can be impactful in implementation even though it is not happening yet. In relation to development and evaluation, Riley deems GenAI helpful as she uses it to generate assessments and suggestions for evaluation criteria. Riley also creates assessment rubrics and pointed out that GenAI makes it efficient. For Riley, GenAI sometimes provides incorrect information thus being a tool that offers suggestions rather than a final product since she needs to edit and check thus thinking that human touch is still needed.

3.2. Factors Affecting Instructional Designers’ Integration of GenAI into Their Professional Practice

3.2.1. Nikki’s Attitudes

Nikki defined GenAI as “specifically dealing with GenAI, which is based on large language models which are trained on an incredible amount of data to be able to recognize in a very human-like nuanced way”. Nikki extended it by stating that GenAI “works with patterns of speech or even visual design in a way that can be leveraged to do a lot of the work that humans traditionally are responsible for doing”.

Nikki had ambivalent attitudes toward GenAI: “So, it’s a bit of a love-hate relationship with it”. While acknowledging GenAI can be useful in ID, Nikki was concerned about the possible negative shift GenAI may bring in people’s everyday lives and careers. Despite this nuanced feeling, Nikki seems to have developed skills and competencies related to GenAI: “…I’ve come to embrace it a lot more. I use it on a daily basis. I see a lot of positives about it”.

Nikki explained that GenAI can also be used as a student resource to support their learning and an instructor guide to map out the objectives and competencies. Nikki described that IDers can create a case study with the narratives and storylines using GenAI. Nikki also acknowledged the substantial impact GenAI can have on people’s lives and education: “I think that it will probably find its way into many of the roles, responses really, into the threads of our lives…next big thing in education”.

Nikki was also positive about future opportunities GenAI will bring while acknowledging potential threats. Emphasizing how AI can be used to automatize some tasks, she indicated: “…so that I can do the things that I really love to do. And, so, I hope that it does create an opportunity to focus on the most human parts of our work”. However, Nikki showed concern over how technology drives procedures and policies highlighting the potential risks of lack of human insights and ethical considerations.

3.2.2. Taylor’s Attitudes

Taylor defined AI as “computer-generated information or knowledge” that is “getting smarter all the time”. Having many years of ID experience, she showed mixed feelings about GenAI and its future. Taylor also expressed concern over potential bias and false information: “you can’t trust that anything you see or hear online is real anymore. It’s generating things and it’s causing just the potential for misuse”. However, Taylor also expressed excitement, curiosity, and anticipation: “I’ll spend a couple hours with this software and see what I can come up with and it does a pretty good job. It feels like it’s improving. So, I’m excited to see where the work goes”.

Taylor perceived GenAI useful at generating narratives and ideas and highlighted that GenAI excels in helping with brainstorming: “That was another kind of feature that I thought could help me, get into it and make my day a little bit more productive and faster”. Taylor also indicated how GenAI can directly change classroom instruction and modeling. Taylor provided an example where teachers can use 3D printers to introduce a product idea and a marketing strategy. Taylor also believed that GenAI would bring innovation.

Regarding current GenAI policies, Taylor felt that it depends on the institution and work environment to make the relevant decisions. Taylor added that restricting the tool may not be helpful: “But I think banning a platform is not gonna solve what their actual fear is”. Taylor also emphasized the efficiency and convenience of GenAI: “If I can get the machine to do something I find tedious, I have more time to do XYZ that I really enjoy doing so maybe we could come to some kind of model where it gets customized to the person really”.

Although Taylor was positive about the future opportunities and impact of GenAI, she pointed out possible threats such as misinformation and disparity. Regarding people’s concern over unemployment, Taylor added: “There needs to be a different system in place to support people who, you know, lose their job through no fault of their own”.

3.2.3. Elvin’s Attitudes

Elvin had ambivalent attitudes toward GenAI and its implications for the future. As a university IDer, his main concern was the possibility of GenAI replacing IDers and threatening the field: “I don’t like it from a professional perspective, because it’s now going to take a while to fully replace somebody like me with AI. Some of the people that I have worked with in the past have already stopped working with me, because they can do it for free using ChatGPT and applications such as that”.

Elvin provided an example of using ChatGPT as a brainstorming tool with a positive attitude. While acknowledging the efficiency the tool may bring, he expressed several limitations: “But there’s still quite a few bugs in it. You need to edit the content, or not maybe edit but at least double check”. As such, he highlighted that humans must cross-check and revise the content created by GenAI: “But, you know, where is the creativity going to be? They’re going to do it from the—strictly the task. You’re giving parameters, but you’re still not going to have that human nuance, element”.

Elvin expected GenAI to improve and be more capable of various tasks in the future. However, he shared concerns over the possible threat ever-developing tools may bring to ID. For example, he explained that a professional ID expert’s knowledge and background may be reduced to reviewing computer-generated work. He added, “I mean, somebody at my level, getting paid what I get paid, shouldn’t have to be doing that”.

3.2.4. Cameron’s Attitudes

Cameron showed ambivalent attitudes too. While maintaining that GenAI causes some anxiety and currently lacks accuracy, Cameron is motivated to try and explore its advantages. According to Cameron, brainstorming, scripting, expanding ideas, and aiding with video and audio functionality are useful aspects of GenAI: “I find that it’s a very helpful tool to kind of just help generate ideas, help explore new ideas and be a starting point. I would not use it as an endpoint, but I use it as a starting point”.

Cameron also explained that GenAI has been prevalently used, such as generating headlines and improving the speed of labor. Additionally, Cameron took attention to how GenAI has raised the necessity to check trustworthiness and ethics: “How much can I trust it? Was it double checked? The other part is like how ethical”. Accordingly, Cameron highlighted reliability and accuracy as limitations thus raising the need to be critical and discerning: “You have to be careful, because like with everything else, it’s not perfect…It’s really more about critical thinking. It’s really more about soft skills where we, you know, AI cannot do that for us”.

Regarding the future of GenAI and policy, Cameron suggested an embracing and positive mindset focusing on supports and adjustments for a better use: “It will get better. I see it’s getting better, because, no matter what, it’s still a learning”. She also emphasized the need to earn GenAI literacy: “My biggest fear is that, again, I’m in all support of AI, but my biggest fear is that it will do the work for you without you understanding what exactly it does. And it’s very difficult to change or to improve if you don’t realize how it works”.

3.2.5. Jordan’s Attitudes

Jordan recognized GenAI as a massive language learning model that works as an input–output system dependent on what is fed into it. Jordan also showed a nuanced feeling, acknowledging the benefits GenAI brings and the stress it may cause: “I don’t know if that’s necessarily a bad thing or a good thing, because our field’s across so many different sectors in so many different ways”.

As an IDer in a corporate setting, Jordan uses GenAI mainly for generating scenarios, structuring lessons, and asking for confirmation accompanied by double-checking: “I’ve used ChatGPT. Sure, I’ve gone and will go ahead and go back to my textbooks, and I’m going to go back to the other resources that I have again reconfirming my approach”. While acknowledging the huge impact GenAI can have on education and designers, Jordan mentioned the lack of accuracy and ethics as potential issues, highlighting the importance of critical thinking: “There is the risk of whether or not information is valid or not which runs a similar threat to GenAI … You have to understand what it is you’re looking for …”. Likewise, Jordan highlighted the need for systematically supporting IDers on ethical use: “I think the responsibility is ultimately going to be on the more seasoned designers to train newer designers on ethics and accountability in all facets”.

As for GenAI’s future, Jordan anticipated that its use would increase as it gets more common among especially the younger generations. However, Jordan showed concerns over IDers’ becoming overly dependent on it, which may deteriorate skills development. He underscored an awareness of work ethic, mental capabilities, and developmental growth: “There’s still that human element to it. But I think I do worry. Areas that the use of GenAI make organizations feel as if instructional design and instructional designers, that human component, is obsolete…”.

3.2.6. Riley’s Attitudes

Riley perceived GenAI as a large language model trained based on large data. As an IDer working in higher education for more than ten years, Riley generally shared a positive attitude toward GenAI’s usefulness and impact on life and work. Riley mentioned that GenAI helps to be resourceful, efficient, and adaptive at work when brainstorming, creating learning objectives, and providing assessments and reading materials even though the product may not be perfect. She also emphasized the usefulness of GenAI when creating alt-texts for images, at which the tool seems to excel more than humans: “It’s not like it gives me perfect set of learning objectives, but it gives me something in less time to give suggestions to faculty in writing learning objectives… So, I ask Gemini to give me an alt text…I can go over that sometimes, but I give it directions. So, it does a better job than I would.”

Riley acknowledged the immense impact GenAI brought: “It’s going to affect everything, yes I believe that eventually it will”. Riley used GenAI for planning trip itineraries and researching various topics of interests: “So it is very impactful and it’s different than social media because again it responds to your queries instead of you going out and you know finding things on your own and filtering through things”.

Riley also emphasized some limitations regarding quality and raised the need to critically check and edit the product generated: “But sometimes it is, of course, it’s not perfect. It may not provide the correct information. We are not there yet, in my opinion”. Riley anticipated that a more powerful tool would appear soon and will impact ID: “We will have to adopt, and AI will be more powerful, easier to use”.

3.3. Cross-Case Analysis

The participants’ integration of GenAI and their ambivalent attitudes toward its evolving role in ID aligned with each other; while talking positively about GenAI due to its efficiency, they also kept mentioning its potential limitations and drawbacks. Specifically, participants’ attitudes toward GenAI are basically ambivalent due to various benefits and usefulness, and concerns about accuracy and ethics. Most participants reported that GenAI can increase efficiency in life and ID tasks. One participant explicitly voiced their high satisfaction with GenAI to prepare travel plans, for instance. As for ID and education, the participants reported that GenAI can be useful in multiple ways: They mostly think that GenAI can help complete tedious work and increase efficiency so that they can spare more time for more critical tasks, which results in a positive attitude.

However, participants also have some concerns that create negative attitudes. One common concern is the automation of tasks and people losing their jobs or positions. One participant with extensive professional experience has deep concerns highlighting that their clients are already switching to GenAI for content generation since it makes it faster and cheaper. Some also have a negative attitude toward the need to double-check GenAI-created material for accuracy since GenAI-created products are not perfect. Another common negative attitude is related to a potential overreliance on GenAI and misusing it in unethical ways. The participants’ experiences using GenAI in their daily lives and professional practice seem to have impacted their ambivalent attitudes since their experiences included both advantages and disadvantages. All these attitudes seem to be linked to the participants’ understanding of GenAI as a computer-based system informed by large data (positive), the efficiency of creating ID products (positive), inaccuracy and the need to double-check the ID product created through GenAI (negative), using GenAI in daily life (positive), and institutional/organizational support (positive or negative).

As for integration of GenAI into ID process, only one participant used GenAI to thematically analyze user interviews and some institutional data (analysis phase). Another participant reported using GenAI for notetaking and checking the notes taken for analysis purposes (instructional planning). In the design phase, all participants integrated GenAI as a brainstorming tool before and/or after creating drafts of content and learning objectives. Namely, participants reported using GenAI as an ideation tool and highlighted that it is better than starting from scratch. All participants used GenAI to create initial working drafts of content including text, quizzes or assessments, rubrics, scenarios, videos, voiceovers, alt-texts, images, chatbots, and short lessons. So, brainstorming through GenAI is related to both design and development. One participant reported using GenAI for aligning objectives with content, and another one used it to confirm whether ID decisions are effective. Consequently, participants use GenAI mostly in design and development where they expect its biggest impact to occur in a way that GenAI connects the design and development phases.

Two participants also indicated that they use GenAI for evaluation purposes by creating evaluation tools including assessments, quizzes, and corresponding instructions, thus combining evaluation and development phases. Further, most participants reported not using GenAI in implementation. One of them thinks that implementation would be impacted the least by GenAI while another participant thinks that implementation would be more prone to human touch. Still, one participant thinks that GenAI can also impact implementation in the future as it gets more sophisticated.

Another factor related to the participants’ attitudes toward GenAI seems to be institutional support. All but one participant reported that their institutions have a positive attitude toward GenAI and integrating it into ID, and these institutions encourage IDers to use GenAI especially for development purposes without using proprietary information. Interestingly, the participant whose institution does not support GenAI a lot thinks that a high level of basic ID competence is still necessary for using GenAI effectively. Lastly, only one participant thinks that their formal ID degrees helped them prepare for GenAI, but helped with being more creative and critical.

Across the six cases, the common tendency is that participants are still exploring various ways of integrating GenAI into their ID practice, and depending on pros and cons of this process, they develop positive or negative attitudes toward GenAI. In turn, such attitudes can further shape their integration of GenAI into professional ID practice. The participants’ attitudes toward GenAI are also related to other factors including their understanding of GenAI, efficiency of using GenAI, quality of products created through GenAI, using GenAI in general life tasks, and workplace support. Figure 1 summarizes the main themes or factors shaping the integration of GenAI into professional ID practice as informed by our sample:

4. Discussion

This multiple-case study investigated six IDers’ integration of GenAI into their professional practice, and the factors that affect this integration. The six study participants think that GenAI is helpful and use it in various ID practices as they relate to design and development phases mostly. The participants have a very positive attitude toward this use since it makes ID a more efficient process, aligning with previous research (e.g., T. Luo et al., 2025). Specifically, participants use GenAI to conduct tedious tasks and brainstorm ideas for creation, which is necessary in ID (e.g., Bolick & da Silva, 2024; Ch’ng, 2023). Likewise, previous research pointed to using GenAI for brainstorming in ID (e.g., Kumar et al., 2024; T. Luo et al., 2025), and this study adds that GenAI-assisted brainstorming starts in the design phase and can continue through the development phase. Interestingly, one participant also indicated that GenAI can also provide insight into content quality thereby helping with some SME tasks, as also mentioned by Kumar et al. (2024).

GenAI saves time, allowing IDers in our sample to focus on prioritized tasks such as creative design that they would enjoy more, which is also pointed out by previous research (e.g., DaCosta & Kinsell, 2024). As Bolick and da Silva (2024) suggested, GenAI contributes much to content development, including scenarios, images, videos, alt-texts, outlines, summaries, evaluation items including rubrics (e.g., Kumar et al., 2024), and questions (e.g., Rezigalla, 2024). From there, IDers can creatively refine and tailor instructional materials to be appropriately customized based on learner needs and goals. Given that instructional development entails technical skills and time, participants’ GenAI use in development and design phases is not surprising. In other words, GenAI can make design and development phases more efficient despite the potential need to double-check the products created. Using GenAI for brainstorming and drafting content can also be regarded as both design and development since drafting can happen in both phases. In addition, some participants also use GenAI as an assistant that helps with creative ideas, selecting instructional strategies, and reviewing. Therefore, GenAI can further increase the efficiency of ID models that are more dynamic such as rapid prototyping. In this sense, the collaboration between GenAI and IDers can turn out to be very useful in terms of letting the latter focus on quality and long-term effectiveness of learning experiences. At this point, ethics, and legal aspects of integrating AI into ID would be another skill and knowledge area for IDers, which should be addressed in ID training as well.

However, limitations also exist in integrating GenAI into ID since the products may not be optimal and need further revision, a point made by previous research too (e.g., Ngo et al., 2024). Specifically, GenAI can be good at generating ideas and content whose accuracy and credibility still needs to be checked; participants think they must check whether the GenAI-created content is accurate, thus requiring IDers’ involvement, as Oteng-Darko et al. (2024) suggested. In line with concerns including ethics highlighted by previous work (e.g., A. Nguyen et al., 2023), GenAI’s role and contribution to ID needs to be approached carefully from an ethical perspective, which can be addressed in ID training. After all, ID is not merely creating instructional content, as human engagement and decision-making are still essential for quality assurance, which also includes ethical decisions that should not violate important things like intellectual property law. Another important thing is being careful about bias and discrimination that may show up in GenAI-created items. All these limitations seem to contribute to participants’ negative attitudes, but the efficiency of using GenAI seems to override them. Overall, the ambivalent attitudes imply that IDers use GenAI but also have hesitations since they are aware that efficiency may come at the expense of inaccuracy.

Participants have not actively adopted GenAI for analysis purposes because of privacy concerns. Even though many institutions encourage IDers to employ GenAI, feeding it with proprietary information is restricted. Analysis includes multiple components ranging from learners to learning contexts that can be unique to specific populations, cultures, and environments. A detailed and informative analysis would be on-site thus including more relevant data collection as well. This way, possible changes in various data areas including learner preferences may be detected more precisely. Therefore, using GenAI to analyze data in analysis and evaluation seems limited. Moreover, current GenAI seems to have minimal contributions in implementation. J. Luo et al. (2024) also reported that IDers make little use of GenAI in implementation and evaluation while they found that it is used commonly in analysis. It may be because human involvement is more needed in implementation when instruction is delivered. Hypothetically though, as GenAI tools get more sophisticated, they may demonstrate better analysis functions, mimic learner characteristics better, and can be used more for analysis and implementation purposes as well.

Most institutions involved in this study are positive about GenAI and its use in ID by employing encouraging policies. Such policies seem to be important for IDers to use GenAI since one participant whose institution does not allow the use of GenAI did not report very positive attitudes. To meet institutional expectations, participants use GenAI, but they also highlighted the importance of IDers’ competencies. Therefore, considering that customized learning should be based on ID and learning theories (Katiyar et al., 2024), human intervention through foundational ID knowledge seems critical in using GenAI. This aligns with Hodges and Kirschner’s (2024) claim that the focus should be on designing instruction that enhances learning. Still, foundational ID knowledge can be strengthened by GenAI literacy and competency implying that ID training can integrate GenAI so that students can have both foundational ID knowledge and become GenAI literate. Therefore, GenAI literacy may become an important instrument for professional ID practice in a way that is informed by ID foundational knowledge and how people learn, which would be provided by quality ID training.

Finally, it is worth stating that the current findings align with (a) the technology acceptance model’s focus on perceived usefulness relating to attitudes which would, in turn, relate to intentions to use and actual use (Davis, 1989), and (b) studies adding institutional barrier as a factor relating to perceived usefulness (e.g., Shin et al., 2022) since institutional support seems to relate to the participants’ attitudes toward GenAI and integration of GenAI into their professional practice. Overall, it seems that the participants’ behavioral intention to integrate GenAI into their ID work is informed by perceived usefulness and positive attitudes created by the perceived usefulness, which results in actual use of GenAI especially for design and development purposes. Likewise, the participating IDers seem to still be adopting GenAI on the job based on favorable performance (efficiency or usefulness) and effort expectancy (ease of use), facilitating conditions (infrastructure including AI technologies), and positive social influence (institutional support) (the Unified Theory of Acceptance and Use of Technology or its extension: (Venkatesh et al., 2003). The finding shows that the participants have concerns about the accuracy of GenAI-created products, which relates to their attitudes toward the technology, and is also in line with previous studies showing that trust relates to perceived usefulness, attitudes, and intention to use AI (e.g., Choung et al., 2023). Overall, instructional designers seem to be in an early stage of integrating GenAI into their professional practice and still exploring how to use it to achieve professional goals.

5. Conclusions

The current findings revealed that IDers still seem to be in an early stage of GenAI acceptance and adoption as it relates to their professional ID practice in corporate and higher education contexts. Given that GenAI has become popular relatively recently, this early stage is not surprising. IDers seem to check pros and cons of integrating GenAI into their professional practice and use it accordingly thereby having ambivalent attitudes toward the technology. These attitudes also seem to be shaped by advantages and disadvantages of using GenAI both on the job and in daily life as well as an organizational approach to GenAI. Accordingly, professional ID training may consider ID candidates’ attitudes toward GenAI and the factors affecting these attitudes.

The design and development phases seem to have been impacted by GenAI the most, since GenAI-created products including visuals and ideation through GenAI seem to take less time and effort, thus saving time that can be spent on more creative and critical tasks. Even though IDers may need to check the accuracy of these products, still, the level of efficiency involved in the process seems to make it attractive for IDers. Similarly, integrating GenAI into evaluation appears to focus on designing and developing evaluation tools. Accordingly, GenAI-created items and their evaluation in relation to foundational ID knowledge and how learning happens can be a more integral part of ID training in the future. Clearly, ethical, and legal aspects of integrating GenAI into ID need to be added into ID training, given that IDers have some accuracy concerns regarding products created through GenAI. All these insights can help both IDers and ID trainers to achieve a more informed and purposeful integration of GenAI into ID, which prioritizes learning itself.

Even though this study provided unique insights into IDers’ attitudes and GenAI integration, results should be approached by paying attention to some limitations. Firstly, there were six participants involved in the study and despite their wide range of experiences and professional contexts, results were limited. Future research should consider including IDers with different educational backgrounds and professional profiles. Second, even though a multiple-case study provides rich insights, the findings would be limited to research context thus entailing further research on IDers’ use of GenAI in different workplaces. Third, the current study employed a mainly inductive coding process; future research might employ a combined deductive and inductive approach using existing theoretical and/or conceptual insights. Finally, due to proprietary concerns, the main research data came from interviews only, and future research would include other data sources including sample ID products created through GenAI.