Promoting Academic Integrity in AI-Practice—The Effect of Live Coaching in Higher Education

Abstract

The rapid spread of generative artificial intelligence (AI) in higher education creates both opportunities for innovation and challenges for academic integrity, ethical use, and students’ critical thinking, particularly in scientific writing. This study examines whether a synchronous live coaching format can support students in developing reflective and responsible AI practices. A mixed-methods cross-sectional evaluation was conducted at a German distance-learning university with a strong focus on health and social sciences. An online survey was administered to 168 students who participated in voluntary live coaching sessions on “AI in Scientific Writing”. Quantitative items assessed perceived competence gains, ethical awareness, and confidence in handling AI tools, while open-ended questions captured qualitative feedback on the format’s strengths and improvement needs. Students reported that the coaching enhanced their understanding of responsible AI use and scientific integrity and valued the opportunity for open discussion, peer interaction, and the supportive attitude of instructors. Reflective and dialogic elements were perceived as particularly beneficial. Overall, the findings suggest that synchronous live coaching can contribute to fostering ethical awareness and higher-order thinking in AI-supported academic work, especially when it integrates structured input with dialogue, reflection, and peer learning.

1. Introduction

The rapid development and proliferation of artificial intelligence (AI) bring both significant opportunities and challenges for innovation in higher education [1]. In view of rapid technological developments, AI has the potential to revolutionize higher education, enriching many processes through its innovations. Nevertheless, the digital divide, digital literacy, and ethical issues remain key challenges in its implementation [1]. These developments are particularly evident in the emergence of applications like ChatGPT, which have sparked widespread public discussion and initiated a profound change in the way AI is approached [2]. Simultaneously, research findings indicate that large language models (LLMs) possess a remarkable level of reasoning abilities, extending across both text-based and visual modalities [3]. Especially in light of these capabilities, the question arises whether traditional examination formats in the academic context require a critical reassessment to adequately address the growing abilities of AI tools such as LLMs [4].

Kamalov et al. [5] demonstrate that AI-supported learning tools can provide substantial advantages for students by facilitating a personalized learning experience, allowing learners to tailor both the pace and methodology to their individual preferences. Students further indicate that AI-supported learning tools enable them to adapt learning materials to their “personal learning style and pace” [1]. At the same time, the integration of such technologies requires adjustments to the curriculum and the provided materials to ensure that content and assignments are optimally aligned with the use of AI and that learning objectives continue to be met [5]. However, students express concern that a strong reliance on AI technologies could potentially limit the development of critical thinking skills [6]. The unrestricted access to AI tools and the ability to individually design one’s learning trajectory can paradoxically induce stress, as students bear full responsibility for organizing their own learning process [5]. This dual aspect—comprising concerns about potential cognitive constraints on one hand and the demands of self-directed learning on the other—highlights an ambivalent stance toward AI-assisted learning, encompassing both opportunities and challenges for academic advancement.

The responsible use of AI poses a significant challenge, particularly in the context of scientific writing. The increasing support provided by digital tools in the scientific process requires that skill development focus on core competencies: the critical evaluation of sources, careful analysis of information, ethical reflection on one’s own methods, and the ability to responsibly situate knowledge within both scientific and societal contexts [7]. The Stiftungsverband, a leading German network involved in advising and shaping education and science, also emphasizes that it is essential to strengthen scientific working methods and to promote the ability for critical and independent research, as well as the traceability of sources [8]. This entails the necessity of considering preventive strategies that higher education institutions can adopt to uphold and foster academic integrity [9] while concurrently assisting students in cultivating their critical and self-directed learning competencies, with a primary focus on the promotion of critical thinking [4]. Higher education institutions are called upon to build AI literacy and implement ethical guidelines for the equitable use of generative AI [10].

The deployment of AI-supported tools presents more limitations and risks than scientific integrity. Gerlich [11] reports that frequent engagement with AI tools correlates with reduced critical thinking abilities. Moreover, the use of AI tools like ChatGPT seems to exacerbate tendencies toward procrastination and diminished memory retention, consequently adversely affecting students’ academic outcomes [12]. Furthermore, the use of AI in education raises concerns regarding data protection, security, potential biases, and the changing dynamics between instructors and learners [5]. This underscores the need to understand AI as a supportive, yet thoughtfully applied, tool in the scientific process.

In response to this context, Apollon University developed an interactive live coaching format dedicated to this subject. In regularly scheduled digital group sessions, the program focuses on ‘AI in Scientific Writing.’ Its objective is to sensitize students to the reflective use of AI tools in practice and to empower them to maintain scientific integrity. The format aligns with higher education pedagogical German guidelines, approaching AI as a tool that necessitates critical reflection, with outputs that are to be scrutinized, questioned, and contextualized in relation to traditional sources [13].

2. Methods

2.1. Subsection

The evaluation was conducted at a distance-learning university with a large and diverse student population of approximately 6500. The university has a strong focus on health and social sciences, particularly in the field of health economics and management. Its largest programs include Nursing Management, Psychology, and Social Work. Among the students, 77% identify as female, 22% as male, and 0.08% as diverse, with an average age of 37.5 years. The age distribution is broad, with the majority of students aged between 20 and 49 years: approximately 30% are 20–29, 38% are 30–39, and 22% are 40–49. Students over 50 account for around 10% of the sample, whereas those under 20 (0.2%) or over 70 (0.1%) represent only a very small fraction.

2.2. Format

The monthly live coaching represents a voluntary low-barrier support format, accessible to all students without participation fees. The structure of the format is intentionally designed to maximize accessibility and usability:

• Sessions are scheduled consistently at the same time each month and are offered in the evening, allowing participation by students who are employed.
• Access credentials remain unchanged across sessions, ensuring seamless participation.
• Students may join or leave at any time, with no prior registration required.
• No minimum or maximum number of participants is prescribed; the session proceeds irrespective of group size.
• Each session is conducted by two instructors, one of whom primarily oversees chat interactions to facilitate questions and discussion alongside the live session.

The coaching is offered at the university as a monthly, 90 min group session for students from all degree programs.

The synchronous format attracts a large number of students, highlighting the strong need for guidance at the intersection of technological progress and ethical responsibility. It builds on best-practice examples from Germany, where students are empowered in moderated live workshops or peer formats to critically reflect on the functions, limitations, and risks of generative AI [14]. The format is further regarded as a low-barrier offering, requiring no registration and available free of charge. The same virtual space is accessible to all students—irrespective of their field of study—at scheduled times each month, thereby actively promoting informal and spontaneous engagement.

The format follows a two-phase structure, combining an input phase that introduces key aspects of scientific work in the context of AI through practical examples with a dialogue phase that addresses students’ questions in depth while maintaining scientific integrity. The learning objectives are embedded in the revised Bloom’s Taxonomy, guiding students from basic knowledge recall, such as understanding core principles of scientific integrity, to higher-order skills including critical analysis of AI-generated texts, ethical evaluation of AI use in different phases of academic work, and the creation of personal guidelines for responsible AI use [15]. The focus is not on the mere use of AI but on critical engagement with scientific sources. Reflection tasks, such as evaluating AI-generated texts or developing personal usage guidelines, promote an integrity-oriented approach [16].

Building on this theoretical framework, a central research question emerges: In what ways does participation in the live coaching format enhance both the ethical awareness and practical competence of students when working with generative AI in academic contexts?

2.3. Study Design

The evaluation of the newly introduced ‘Live Coaching’ format was conducted as a one-time cross-sectional online survey. Participants were provided with a link to the voluntary and anonymous evaluation during a regular live coaching session. No personal data were collected that would have allowed the identification of individual participants.

Data collection was carried out using a standardized online questionnaire. The questionnaire consisted of five thematic sections and was designed to capture students’ subjective assessments of the extent to which the session contributed to their academic development:

• General information about the students
  This section captured general characteristics (degree program, study progress, frequency of AI use), as well as information on previous participation in the live coaching sessions.
• Perceived effectiveness of the live coaching
  This section assessed students’ subjective evaluations of the format’s effectiveness, focusing on the following:
  • Competent and responsible use of AI;
  • Ethical and legal aspects;
  • Scientific integrity;
  • Perceived competence gains resulting from the live coaching.
  These items align with core competencies emphasized in AI-supported scientific work, including the critical evaluation of sources, ethical reflection, and responsible knowledge application [7,8,10]. Responses were measured using four-point Likert scales ranging from ‘strongly disagree’ to ‘strongly agree’.
• Competence in dealing with AI outputs
  Students assessed how likely they are to critically evaluate, verify, and responsibly use AI-generated results in their studies. This measure was included to capture the perceived ability to translate AI literacy into practice, addressing concerns that AI use can affect critical thinking, academic integrity, and study outcomes [11,12]. Responses were measured using a four-point Likert scale ranging from ‘very unlikely’ to ‘very likely’.
• Open-ended questions
  In addition, the questionnaire included several open-ended items allowing for qualitative assessment. These questions did not specifically address scientific integrity or ethics, but were formulated openly:
  • What did you find particularly helpful in the consultation session?
  • What would you like to see in future consultation sessions (content, methods, format)?
  • Any further comments or feedback?
  These items were included to capture students’ reflections, suggestions, and experiences in their own words, providing qualitative insights to complement the quantitative measures of AI competence and perceived effectiveness.

2.4. Implementation and Data Processing

The survey was conducted online through an internal university survey platform. Participation was voluntary via a non-personalized link distributed during the live coaching session, and the data collection was fully anonymous.

Quantitative data were analyzed descriptively, including means, standard deviations, and frequencies. Open-ended responses were subjected to a structured qualitative content analysis for coding and interpretation.

3. Results

3.1. General Information

A total of 168 students participated in the survey. The majority were enrolled in Bachelor’s degree programs, with a smaller proportion in Master’s degree programs. Students were relatively evenly distributed across the various stages of study, with the smallest proportion being those prior to or in the midst of their thesis. Concerning the use of AI tools, most students indicated occasional usage, a smaller fraction reported regular usage, and roughly one-third currently do not use AI. Many students were participating in the live coaching for the first time; however, some indicated attending multiple sessions or participating on a regular basis. A detailed description of the sample is provided in

Table 1. Sample description (N = 168).

Variable	Category	N (%)
Course of Study	Bachelor’s	132 (78.6%)
	Master’s	33 (19.6%)
	No response	3 (1.8%)
Phase of study	Early stage of study	73 (43.5%)
	Prior to/midst to term paper	57 (33.9%)
	Prior to/midst of thesis	37 (22.0%)
	No response	1 (0.6%)
Use of AI tools	No	61 (36.3%)
	Occasionally	85 (50.6%)
	Regularly	21(12.5%)
	No response	1 (0.6%)
Frequency of participation	Once	77 (45.8%)
	Multiple times	65 (38.7%)
	Regularly	22 (13.1%)
	No response	4 (2.4%)

.

3.2. Perceived Effectiveness of the Live Coaching

The respondents rated different aspects of the session overall positively, and answers were given on a four-point scale. The mean score of 3.46 (SD = 0.87) suggests that students’ understanding of responsible AI use was enhanced. Similarly, awareness of scientific integrity was reported as considerably improved, with a mean of 3.45 (SD = 0.73). Students’ confidence in handling AI tools within the context of scientific work was rated at 3.05 (SD = 0.91), whereas their ability to evaluate the ethical appropriateness of AI tool usage in their studies received a rating of 3.15 (SD = 0.73). The opportunity to openly address questions or uncertainties related to AI use was considered especially valuable, with a mean rating of 3.46 (SD = 0.81). Peer interaction during the session was rated 3.30 (SD = 0.82) on a four-point scale and was considered beneficial for the learning process in scientific writing. The session also enhanced participants’ confidence in applying citation rules (M = 3.25, SD = 0.85). Overall, the findings reflect largely strong agreement, with the moderate dispersion of responses indicating a fairly uniform perception of the content.

3.3. Competence in Dealing with AI Outputs

Figure 1 depicts students’ competence in handling AI outputs, assessed as the third section of the survey. The findings suggest that students anticipate critically evaluating AI-generated outputs prior to their use. Furthermore, they findings show that students would employ AI applications in a transparent manner, including documenting them in the university’s mandatory approved tools list. Students are also likely to encourage peers to engage with AI tools reflectively and are expected to apply these tools in a reflective and accountable way in their own academic work.

3.4. Open-Ended Questions

The evaluation of the live coaching sessions focused on three key areas: students’ perceptions of the session’s helpfulness, suggestions for future content, methods, and format, and any additional comments or feedback. The following section presents the main findings from each of these areas, highlighting both the quantitative ratings and qualitative insights provided by participants.

3.5. Most Helpful Aspects of the Session

3.5.1. Peer Interaction

One of the most prominent themes reported by students was peer interaction. Many participants noted that hearing the questions of other students was particularly helpful and reinforced the sense of not being alone. Illustrative comments included the following: “The open exchange and the feeling that I am not alone with my problems” and “The interaction—realizing that other students are experiencing the same challenges as I am”. The value of having a structured exchange session was frequently emphasized, with a notable subtheme being that such interactions foster a perceived “feeling of security”.

3.5.2. Respectful Interaction

Another recurring theme concerned the supportive atmosphere of the live coaching sessions and the demeanor of the instructors. Students commented on “an open and respectful interaction among participants”, “the wonderfully relaxed and warm manner…”, and “very competent, authentic, motivating, and cheerful approach”. One participant added: “The warm and welcoming manner makes it very easy to engage and accept the guidance”.

3.5.3. Responsible AI Usage

Students also highlighted both the perceived threshold and the practical concrete use of AI tools. Many appreciated “an honest engagement with the topic of AI” and the “motivation that AI is now an integral part of the process and can be used effectively when applied with care”. Responses emphasized reflective use, for example: “AI is not fundamentally excluded but is adopted over time, with a respectful and critical approach being demonstrated”. Students valued practical guidance on AI usage, illustrated by comments such as “AI as a sparring partner” and “the use and evaluation of search terms, employing AI as a sparring partner”. At the same time, some students noted hesitancy in using AI, stating: “AI use is presented neither as the method of choice nor as something to be feared”, and “I am still so uncertain about it that I simply avoid using AI altogether”.

3.5.4. Further Comments and Feedback

Students provided additional suggestions and reflections to further improve the live coaching sessions. Several participants highlighted the usefulness of the format itself, with one noting: “Very helpful format”. Suggestions for enhancement included collecting questions in the chat for structured responses: “Collect questions in the chat and use them for answering”, offering sessions more frequently: “More frequent sessions”, and providing recordings for later reference: “Gladly record them!” Overall, the comments indicate that students value the live coaching sessions and are interested in measures to increase accessibility, structure, and frequency.

4. Discussion

4.1. Frequency of AI Usage

As shown in Table 1, a substantial proportion of students reported not using any AI tools. Compared to other universities, this group appears relatively inexperienced with AI, which may influence their learning needs and the types of support they find most helpful [17,18]. This suggests that the student population in the present course may have less prior experience with AI, which could influence both their learning needs and the types of support they find most helpful. It should also be noted that large parts of the student population are already working in health-related professions, particularly in nursing. This professional background may further shape their experiences with AI and their specific learning needs.

4.2. Phase of Study

Table 1 shows that the majority of participants are Bachelor’s degree students, while a smaller group are Master’s degree students. Students in earlier stages of their studies generally bring more general questions about AI and scientific writing, whereas those further along may have more advanced or specialized inquiries. In the open-ended responses, several participants suggested that live coaching sessions could be tailored according to degree program or study phase to better address these differing needs. Offering different formats tailored to students’ degree programs and progression stages could therefore be beneficial. However, a key theme emerging from the open-ended responses was the value of peer interaction. Having students with diverse skill levels in the virtual session can support those who feel less confident or have less prior knowledge, creating opportunities for collaborative learning. Nadile et al. [19] underscore this issue, noting that fear often prevents students from asking questions in challenging classes, and that observing peers ask questions can be particularly helpful, especially for first-generation college students. It is likely that many students at the evaluated university are first-generation students, highlighting the importance of peer support.

4.3. Effectiveness

Overall, the format was rated positively across all assessed topics. Students indicated that the live coaching effectively promotes the understanding of responsible AI use and provides opportunities for open discussion of questions or uncertainties. These elements promote higher-order thinking in line with the revised Bloom’s Taxonomy [15], addressing skills such as analyzing AI-generated texts, evaluating ethical implications, and creating personal strategies for responsible AI use, which are crucial competencies in the AI era.

Students particularly valued the reflective and discussion-based aspects of the coaching. These results indicate that students particularly valued the reflective and discussion-based aspects of the live coaching, complementing the more technical or skill-oriented components of AI instruction. This finding is consistent with previous research emphasizing the importance of fostering ethical engagement and digital literacy in higher education [5,17]. The open discussion was also a large topic in the qualitative part of questioning, where students made very clear that the attitude of the teachers was an important factor in the success of the format. Cureton et al. [20] also emphasized the importance of the “human touch” in higher education—an aspect to which distance-learning institutions must pay particular attention. Bell [21] similarly notes that students highly value respectful and warm interactions with their instructors. Approachability and empathy are frequently cited as important qualities, a finding that is also reflected in the results of this evaluation.

The highest consensus regarding the effectiveness of the format was observed for raising awareness of scientific integrity and evaluating the ethical appropriateness of AI use. This indicates that students largely agreed on the suitability of the format for these topics and that students are engaging in higher-order thinking at the ‘Evaluate’ level of Bloom’s Taxonomy [15]. In the AI era, it is particularly important to deliberately target higher-order cognitive skills, since it is indicated that students revert to lower-level thinking (Remembering and Understanding) when faced with complex or uncertain AI-supported tasks [22]. Raising awareness of scientific integrity issues in AI use and teaching responsible engagement is a major concern in academia [23] and was a primary goal of the live coaching. The results suggest that students feel supported in this regard, highlighting the effectiveness and importance of such an offer. These findings are consistent with the responses, where responsible AI usage was frequently mentioned as a helpful aspect of the live coaching.

The area of greatest uncertainty among students was the safe use of AI tools in the context of scientific writing. Students expressed lower agreement and more heterogeneous opinions in this domain, indicating a need for further guidance and support to build confidence in responsible and secure AI usage.

Students particularly valued the reflective and discussion-based aspects of the coaching. These results indicate that students appreciated opportunities to engage in higher-order thinking, aligning with the upper levels of Bloom’s revised taxonomy [15]. In the AI era, it is especially important to target these higher-order cognitive skills, such as analyzing AI-generated texts, evaluating the ethical implications of AI use, and creating personal strategies for responsible AI engagement. Addressing these advanced cognitive processes helps ensure that learning objectives go beyond basic knowledge recall and truly prepare students for ethical and responsible AI usage in scientific work.

5. Conclusions

This evaluation highlights the value and relevance of a live coaching format designed to support students in the responsible use of generative AI in academic contexts. Students at the participating institution reported comparatively limited prior experience with AI, underscoring the need for structured guidance and accessible learning opportunities. The results suggest potential differences in students’ needs depending on their study phase, indicating that offering differentiated formats could further enhance the effectiveness of such interventions. At the same time, peer interaction emerged as a central benefit, particularly for students with lower confidence or less prior knowledge.

Students emphasized the importance of reflective engagement with AI, open discussion of uncertainties, and the supportive stance of instructors. These elements were consistently identified as key strengths of the format and align with broader calls in higher education to foster ethical, responsible, and informed AI use. The strong consensus regarding topics such as scientific integrity and ethical evaluation further demonstrates that students perceive these areas as well supported by the coaching.

However, uncertainty remains around the safe use of AI in scientific writing, indicating that students need additional guidance to navigate risks and limitations. First, this cross-sectional design relied primarily on students’ self-reported perceptions following participation in the live coaching sessions. The absence of a control or comparison group limits the ability to attribute observed effects solely to the intervention.

Second, the study was conducted at a single distance-learning university, with a sample largely composed of students from health and social science disciplines and a relatively high average age. These characteristics restrict the generalizability of the findings to other institutional contexts or more traditional student populations. The findings may not generalize to other types of universities, different fields of study, or students with more prior experience using AI.

Third, although the results indicate high levels of student satisfaction and perceived usefulness of the live coaching format, it remains unclear whether these perceptions translate into actual changes in academic behavior, such as citation practices or transparent disclosure of AI use.

Given these limitations, the present study should be considered exploratory. Future research could strengthen the evidence base by employing validated instruments, longitudinal or controlled study designs, and samples that include diverse disciplines, institutional contexts, and student demographics. Future research could investigate whether the effectiveness of live coaching differs between students with high prior AI literacy and those new to generative tools and whether interventions could be tailored based on experience level or field of study. Additionally, research could explore how AI support can be most effectively structured, potentially tailoring interventions based on field of study, prior experience, and learning format. Overall, the findings suggest that live coaching, combining clear instructional guidance with opportunities for dialogue, reflection, and peer learning may play a meaningful role in promoting AI literacy and ethical awareness, but further investigation is needed to confirm these effects in practice.