Turning the Page: Pre-Class AI-Generated Podcasts Improve Student Outcomes in Ecology and Environmental Biology

Abstract

In the aftermath of the COVID-19 pandemic, instructors in higher education have reported a decline in foundational reading habits, particularly in STEM courses where dense, technical texts are common. This study examines a low-barrier instructional intervention that used generative AI (GenAI) to support pre-class preparation in two upper-division biology courses. Weekly AI-generated audio overviews—“podcasts”—were paired with timed, textbook-based online quizzes. These tools were not intended to replace reading, but to scaffold engagement, reduce preparation anxiety, and promote early familiarity with course content. We analyzed student engagement, perceptions, and performance using pre/post surveys, quiz scores, and exam outcomes. Students reported that the podcasts helped manage time constraints, improved their readiness for lecture, and increased their motivation to read. Those who consistently completed the quizzes performed significantly better on closed-book, in-class exams and earned higher final course grades. Our findings suggest that GenAI tools, when integrated intentionally, can reintroduce structured learning behaviors in post-pandemic classrooms. By meeting students where they are—without compromising cognitive rigor—audio-based scaffolds may offer inclusive, scalable strategies for improving academic performance and reengaging students with scientific content in an increasingly attention-fragmented educational landscape.

1. Introduction

The COVID-19 pandemic exacerbated existing trends in higher education around student disengagement from traditional reading assignments (Kinzie, 2023). Interruptions to in-person instruction, shifts to virtual learning, and increased reliance on digital media have collectively weakened foundational reading habits, particularly among students from under-resourced communities (Armstrong et al., 2022). Library data also reflect this broader shift, with declining print book circulation suggesting a long-term pivot away from traditional academic reading formats (Thornton et al., 2024). This disruption has contributed to declining stamina for sustained academic reading, growing dependence on multitasking, and shortened attention spans—factors that impact not only comprehension but also self-regulated learning in science education (Bone et al., 2025). These trends occur amid widespread adoption of GenAI tools; for instance, over 90% of students report using ChatGPT (OpenAI GPT-4o) for summarization and idea generation—even without formal guidance—indicating that academic preparation practices are being fundamentally reshaped (Ravšelj et al., 2025).

Compounding these challenges is a pedagogical mismatch: while instructors continue to assign textbook-based readings that require close engagement, students increasingly rely on platforms and modalities that offer more passive or abbreviated formats. Generative artificial intelligence (GenAI) tools, such as ChatGPT and other large language models, now provide fast, automated summaries of textbook content, often at the expense of depth and context (McMurtrie, 2024). Although these tools can support surface-level comprehension, their effectiveness as pre-instructional scaffolds in STEM (Science, Technology, Engineering, and Mathematics) disciplines remains poorly understood. Moreover, the rapid proliferation of GenAI tools presents both promising possibilities and complex challenges for educators aiming to preserve academic rigor and ensure equitable access. Recent meta-analytic evidence suggests that while ChatGPT can improve learning performance, perceptions, and higher-order thinking, these benefits are highly dependent on instructional design and context, reinforcing the need for carefully structured integrations (Wang & Fan, 2025). Research highlights the risk of diminished learning quality if AI is adopted without thoughtful pedagogy (Giannakos et al., 2025), and underscores how superficial reliance on GenAI-generated content can inadvertently lower academic standards (Michel-Villarreal et al., 2023).

This study responds to this evolving educational landscape by piloting a low-barrier instructional intervention of short, chapter-aligned audio overviews—hereafter referred to as “podcasts”—generated using GenAI and integrated into the learning workflow of two upper-division biology courses at a Hispanic-Serving Institution. These biology courses were lecture-based and content-intensive, with weekly textbook readings forming a foundational component of instructional design. The podcasts were paired with timed, online pre-lecture quizzes derived from textbook content. Students were expected to listen to the podcast and complete the quiz prior to attending lectures, with the goal of supporting content engagement, easing preparation anxiety, and reinforcing active learning habits.

Importantly, these audio overviews were not positioned as a replacement for reading but rather as a complementary support designed to reduce the cognitive and emotional barriers students experience when confronting dense scientific material. Prior studies have shown that replacing or supplementing traditional lectures with audio-based content such as podcasts can improve learning outcomes (O’Bannon et al., 2011), and by providing students with GenAI-generated scaffolding aligned to each chapter, this approach supports self-regulated learning strategies such as goal-setting, time management, and metacognitive awareness—factors known to promote persistence and engagement in STEM disciplines (García-Peñalvo, 2021; Panadero, 2017). While prior studies have demonstrated the effectiveness of instructor-created audio summaries (Enríquez et al., 2023; Wakefield et al., 2023), recent evidence shows that personalized AI-generated podcasts can further enhance learner engagement and content retention (Do et al., 2025). The emphasis was on re-engaging students in meaningful, preparatory learning behaviors through multimodal scaffolding aligned with core course content, while lowering the entry barrier for students who may lack prior exposure to disciplinary reading strategies. Emerging research finds that repurposing science podcasts as data for AI models enhances understanding of STEM content across multiple subject domains, suggesting that audio may represent a pedagogically rich medium for AI-mediated learning scaffolds (Jia et al., 2025).

The literature on GenAI tools in education has expanded rapidly, particularly in the wake of the release of ChatGPT in 2022 (Qian, 2025). However, few studies have examined how these tools function in STEM learning environments as structured pre-instructional interventions, particularly in biology courses where dense terminology and conceptual layering demand robust comprehension. Prior research has explored the integration of instructor-created podcasts and video-based content to enhance student learning (Enríquez et al., 2023; Wakefield et al., 2023). More recently, studies using personalized AI-generated podcasts—such as PAIGE—demonstrate improved engagement and academic outcomes compared to traditional reading (Do et al., 2025). Nevertheless, there remains a need for empirical evaluations of scalable AI-based scaffolds in upper-division STEM courses.

Numerous studies document how undergraduates in STEM may struggle with developing a robust science identity and sustaining academic motivation—factors that inclusive and scaffolded learning environments can positively influence. Active learning strategies, in particular, have been shown to improve outcomes for all students while narrowing achievement gaps for historically underrepresented groups in STEM (Theobald et al., 2020). For instance, structured STEM intervention programs have been shown to elevate students’ sense of belonging and science identity (Dunbar-Wallis et al., 2024; Shortlidge et al., 2024). Similarly, longitudinal comparisons reveal that even non-STEM students can foster stronger science self-concepts through targeted coursework (Lucas & Vandergon, 2024), while research training contexts reaffirm the power of recognition and meaningful engagement in reinforcing identity development (Pfeifer et al., 2024). Institutional case studies, such as those from IIT Delhi, indicate that GenAI tools are already widely adopted by both students and faculty. These findings underscore the urgency for ethical, transparent, and equitable frameworks to guide integration of AI into academic practice (Dhulia, 2025). The present study positions GenAI-generated audio content as a potentially inclusive strategy for fostering engagement, motivation, and comprehension in content-heavy courses, particularly when paired with quiz-based accountability and aligned course materials.

Against this backdrop, the broader research question guiding this work is this: How can generative artificial intelligence be leveraged to support foundational learning processes in STEM education, particularly in content-intensive courses where students must engage deeply with theoretical material prior to application? This question centers on whether GenAI-supported scaffolds can enhance students’ preparedness, confidence, and engagement with core course content without displacing the development of essential academic skills such as sustained reading and conceptual integration. Within this broader framing, the present study investigated whether AI-generated podcasts could enhance student readiness and academic performance in upper-division biology courses. It examined how students engaged with a GenAI-generated podcast and quiz system embedded into weekly pre-class preparation; what effects this system had on student perceptions of reading, confidence with textbook content, and classroom participation; and whether there were observable patterns in academic performance—such as exam outcomes, quiz scores, and final grades—associated with sustained engagement in the intervention.

We hypothesized that students who consistently engaged with the podcast and quiz intervention would demonstrate higher rates of pre-class preparation, as indicated by quiz completion and performance. We also expected increased self-reported engagement and greater confidence in handling textbook content. Furthermore, we anticipated improved performance on closed-book, in-class examinations and, ultimately, higher final course grades reflective of cumulative engagement and learning gains.

2. Materials and Methods

2.1. Course Context and Participants

This study was implemented during the Spring 2025 semester at the University of La Verne and involved students enrolled in two upper-division undergraduate biology courses: BIOL 323: Ecology (n = 16) and BIOL 312: Environmental Biology (n = 29). Both courses serve as integral components of the biology major curriculum but differ in structure, pedagogical focus, and sequencing. BIOL 323 Ecology is a 2-unit, lecture-based course that introduces core concepts in population, community, and systems ecology. It serves as a prerequisite for upper-division laboratory courses and is typically taken in the second or third year of the major. The course emphasizes theory-driven frameworks and quantitative ecological models. The assigned text was Ecology: Concepts and Applications (Sher & Molles, 2022). BIOL 312 Environmental Biology is a 4-unit course with both lecture and laboratory components, designed to provide students with experiential field-based training in urban ecology, sustainability science, and environmental policy. Students enrolled in this course typically have prior exposure to ecological theory through BIOL 323, though instructional background may vary due to differing faculty, pedagogical approaches, and previously assigned texts. The course used The Biology of Urban Environments (James, 2018). Both courses are required or strongly recommended for biology majors with interests in ecology, environmental science, or applied conservation. Students represented a diverse academic background and varying levels of experience with upper-division coursework and independent learning strategies.

2.2. Intervention Design: AI-Generated Podcasts

To promote engagement with textbook materials and improve pre-class preparation, weekly audio overviews were developed using NotebookLM (free public-facing experimental release available in early 2025), a GenAI tool developed by Google. For each assigned textbook chapter, a PDF of the chapter was uploaded into NotebookLM, which produced an AI-generated spoken summary in WAV audio format. Each audio file was reviewed in full to verify completeness, coherence, and factual accuracy. Recordings that exhibited truncation, distortion, or conceptual inaccuracies were excluded from use. Following quality control, accepted WAV files were converted into MKV video format using VLC Media Player (version 3.0.21; VideoLAN), an open-source multimedia utility. This conversion step was necessary to enable integration with Microsoft Stream, a platform used to distribute the podcast files and embed interactive assessments. Within Microsoft Stream, each video was linked to a corresponding Microsoft Forms quiz, which was programmatically embedded to appear at the conclusion of the audio overview. Each quiz contained 10 multiple-choice questions designed to assess textbook-based content only. No quiz items were derived from podcast-specific wording or phrasing, ensuring that the podcasts served as scaffolding tools rather than replacements for required readings. All podcast-quiz modules were made accessible to students on the first day of the semester. However, to preserve the pre-class preparatory function of the intervention, each quiz was configured to automatically close one hour before the associated lecture. Quizzes could not be reopened post-deadline, and no retroactive participation credit was granted. All submissions were automatically timestamped via Microsoft Forms and recorded in the course gradebook.

To illustrate the integration of podcast and quiz materials within the course infrastructure, Figure 1 presents screenshots of the student-facing digital interface. Panel A displays the Microsoft OneNote Class Notebook, in which each textbook chapter includes a podcast embedded via Microsoft Stream. Closed captioning and a transcript are provided to enhance accessibility. Panel B shows the Microsoft Forms quiz that launches automatically at the end of the podcast. Each quiz is chapter-specific and auto-graded, and delivers immediate feedback. This layout was consistent across all chapters and provided students with multimodal entry points to engage with required readings prior to class.

2.3. Course Structure and Alignment

The BIOL 323 Ecology course was structured as a lecture-only, two-unit upper-division class emphasizing quantitative reasoning and conceptual synthesis. Instruction was supported by weekly lectures integrating foundational ecological principles with field-based case studies and data interpretation exercises. Students completed one immersive virtual reality (VR) assignment designed to examine local ecological impacts of climate change using photodocumentation and spatial scale analysis (Carmona-Galindo et al., 2025). The BIOL 312 Environmental Biology course was a four-unit combined lecture and laboratory course focused on urban ecological systems, biodiversity, and environmental sustainability. Students engaged in a semester-long integrative research project that included site selection, community-based field sampling, biodiversity assessment, and analysis of ecological patterns across biotic and abiotic gradients. The project culminated in the creation and presentation of scientific posters, providing opportunities for authentic scientific communication. In both courses, performance on podcast-aligned quizzes contributed to the participation component of the final grade. These quizzes were one element within a multi-modal assessment framework that also included midterm examinations, cumulative final exams, and writing-based assignments such as data analysis reports and research abstracts. Course content, assignments, and assessments were aligned with upper-division biology learning outcomes, emphasizing critical thinking, data literacy, and communication in the context of ecological and environmental inquiry.

2.4. Pre- and Post-Intervention Surveys

All students were invited to complete pre- and post-surveys designed in collaboration with a doctoral student in Organizational Leadership. These surveys aimed to assess (a) students’ baseline study behaviors and perceptions about textbook reading and (b) changes in confidence, comprehension, and engagement after the intervention. Pre-surveys assessed student familiarity with podcast learning, frequency of textbook use, perceived obstacles to reading (e.g., time management, attention span), and preferred learning modalities. Post-surveys focused on student reflections after using the podcast–quiz system. Likert-style items assessed frequency of podcast listening, perceived effectiveness, engagement, and likelihood of continued use in future STEM courses. Open-ended responses provided additional insight into how students experienced the intervention in both academic and personal terms. All survey data were anonymized and administered digitally via Microsoft Forms. Appendix A provides a side-by-side comparison of pre- and post-survey instruments.

2.5. Data Analysis

Pre-intervention survey responses were analyzed using a mixed-methods approach. Closed-ended items were summarized using descriptive statistics to characterize baseline patterns in students’ study behaviors, textbook reading frequency, time management practices, and preferred learning modalities. Open-ended responses were subjected to thematic analysis, employing inductive coding to identify recurrent themes related to barriers to textbook engagement, such as motivation, scheduling constraints, and multitasking habits. Particular attention was paid to students’ attitudes toward pre-class preparation and their initial interest in audio-based learning tools.

Post-intervention survey data were analyzed to evaluate students’ perceptions of the podcast–quiz system after a full semester of implementation. Quantitative analysis of Likert-scale items focused on identifying usage trends, perceived effectiveness of the podcasts, and self-reported changes in comprehension, confidence, and engagement with course content. Qualitative responses were thematically coded using a grounded approach to capture emergent insights related to learning experiences, accessibility, and the affective impact of the intervention. These qualitative data provided interpretive context for the observed quantitative patterns and informed triangulation of findings across data types.

Podcast-linked quiz performance was tracked using Microsoft Forms, which automatically recorded both participation and individual scores. These data were exported and aggregated for each student to derive two variables: (1) quiz completion percentage (proportion of total assigned quizzes completed), and (2) mean quiz score (average score across completed quizzes). These engagement metrics were then examined for associations with academic performance indicators, including individual exam scores, cumulative exam average, lecture notebook grades, lab notebook grades (if applicable), overall course grade, and attendance.

Pearson’s product-moment correlation coefficients were used to assess the relationships between podcast engagement and each performance variable. Analyses were conducted separately for the Ecology and Environmental Biology courses to account for differences in course design, grading structure, and student cohorts. All analyses were performed using the software package Statistica (TIBCO statistica, 2017).

3. Results

3.1. Pre-Course Survey Responses

Analysis of the pre-course survey (n = 45) revealed limited engagement with assigned textbook readings across both courses. A majority of students reported completing readings only “sometimes” or “often,” with less than 15% indicating that they read all assignments consistently. Ecology students, in particular, demonstrated lower reading compliance: 38.5% reported “rarely” or “never” completing assigned readings, compared to 17.4% in Environmental Biology. These patterns corresponded with reported confidence in comprehension, where most responses fell between “neutral” and “somewhat confident,” and only a small fraction of students expressed strong confidence in their grasp of textbook material prior to lecture. The most frequently reported barriers to completing assigned readings were limited time availability and difficulty maintaining focus, with lack of motivation also cited frequently. Students identified multiple competing priorities, including work obligations and mental fatigue, as contributing factors. When asked about preferred preparatory tools, students in both courses expressed a clear preference for non-text-based modalities, including video content, podcast-style audio resources, and in-class discussions. Despite low reading compliance, students reported moderately high confidence in their overall STEM abilities. However, several retention risk factors emerged, including concerns over workload management, conceptual difficulty, and inadequate academic support systems. Initial attitudes toward the use of podcasts as preparatory tools were generally positive. Many students indicated that audio-based content would better accommodate their learning preferences, particularly if accessible during routine multitasking tasks such as commuting or household chores.

3.2. Post-Course Survey Responses

Post-course survey responses revealed consistent patterns across multiple dimensions of student engagement, confidence, and perceived effectiveness of the podcast intervention (

Table 1. Summary of Post-Course Survey Responses.

Dimension	Survey Indicator	Ecology	Environmental Biology
Podcast Use Frequency	Students reporting podcast use “Often” or “Always”	69.2% of respondents; 46.2% “Always”	69.2% of respondents; 34.8% “Always”
Perceived Comprehension	Podcasts improved understanding of course material	>80% agreement	>80% agreement
Preparedness for Lecture	Reported increased preparedness after podcast use	Majority agreement	Majority agreement
Confidence with Content	Self-reported confidence post-intervention	Primarily “Somewhat confident”; none “Very confident”	Shift toward higher confidence; 21.7% “Very confident”
Engagement Level	Podcasts rated as engaging	More variable; 30.8% “Somewhat engaging”	Predominantly “Very engaging”
Perceived Effectiveness	Podcasts effective for understanding biology concepts	Primarily “Somewhat” or “Very effective”	Primarily “Somewhat” or “Very effective”
Learning Modality Preference	Audio format supported learning	Preferred as a supplement	Preferred as a supplement
Replacement of Textbook	Support for replacing textbook reading with podcasts	Largely unsupported	Largely unsupported
STEM Confidence & Anxiety	Changes in STEM confidence or anxiety	Moderate increase in confidence	Greater reduction in anxiety and improved access

). Podcast use remained high throughout the semester, with over two-thirds of respondents in both courses reporting that they listened “often” or “always.” Ecology students reported higher rates of consistent use than Environmental Biology students, although both cohorts demonstrated strong engagement with the audio content. Across courses, more than 80% of respondents indicated that the podcasts improved their understanding of course material and increased preparedness for lecture.

3.3. Correlations with Course Performance

In the Ecology lecture course (n = 18 students), podcast quiz completion was strongly and positively correlated with overall course grade (r = 0.8055, p < 0.0001). Completion rates also showed significant positive correlations with individual exam grades—Exam 1 (r = 0.5375, p = 0.0214), Exam 2 (r = 0.5068, p = 0.0318), and Exam 3 (r = 0.5843, p = 0.0109)—as well as the average exam score (r = 0.5685, p = 0.0138). Additional positive correlations were observed with lecture attendance (r = 0.6599, p = 0.0029), class notebook grades (r = 0.7508, p = 0.0003), and participation in extra-credit assignments (r = 0.4930, p = 0.0376).

In the Environmental Biology CUREs course (n = 29 students), podcast quiz completion was also positively associated with multiple performance metrics. Strong correlations were found with overall course grade (r = 0.7793, p < 0.0001), lecture portion grade (r = 0.8471, p < 0.0001), lecture notebook (r = 0.5721, p = 0.0012), lab notebook (r = 0.4594, p = 0.0012), lab portion grade (r = 0.4384, p = 0.0174), and lecture attendance (r = 0.4889, p = 0.0071). When podcast scores were graded (rather than completion only), the pattern of positive correlations remained consistent. Graded podcast scores were significantly associated with overall course grade (r = 0.7763, p < 0.0001), lecture portion grade (r = 0.8528, p < 0.0001), lecture notebook (r = 0.5584, p = 0.0016), lab notebook (r = 0.4382, p = 0.0174), lab portion grade (r = 0.4324, p = 0.0191), and lecture attendance (r = 0.4927, p = 0.0066).

3.4. Instructor Observations of In-Class Engagement

In addition to survey responses and performance-based outcomes, the instructor documented descriptive observations of student engagement throughout the semester across both courses. These observations were qualitative in nature and reflect consistent patterns noted during lectures, discussions, and student–instructor interactions. Following implementation of the podcast–quiz scaffold, students demonstrated increased responsiveness to in-class prompts, greater familiarity with discipline-specific terminology, and a higher frequency of content-focused questions earlier in the instructional sequence. Classroom discussions more often moved beyond definitional clarification toward application and synthesis of concepts introduced in pre-class materials. Patterns in student use of office hours also shifted, with visits increasingly focused on refining understanding and integrating concepts rather than initial exposure to course content. These observational data indicate measurable changes in classroom engagement and learning behaviors that align with the survey-based and performance-related outcomes reported above.

4. Discussion

This study examined the implementation of AI-generated podcast overviews as a pre-instructional scaffold in two upper-division biology courses and evaluated their impact on students’ textbook engagement, cognitive preparedness, and overall academic performance. The findings support the hypothesis that pairing GenAI-based audio content with quiz-based accountability can foster more consistent preparation and stronger academic outcomes (Fiorella & Mayer, 2015), while also improving student perceptions of textbook reading.

The pre-course survey data confirmed several barriers that have been widely reported across postsecondary education: difficulty concentrating on dense academic texts, time constraints, and a tendency to rely on surface-level summaries in place of deep engagement (Abbas et al., 2024). These findings are consistent with broader observations about post-pandemic shifts in reading habits (Bone et al., 2025; Mizrachi & Salaz, 2022; Pilotti et al., 2023). Students increasingly express discomfort with traditional textbook assignments and report limited reading fluency, particularly in STEM fields where conceptual layering and discipline-specific vocabulary demand sustained attention. Furthermore, research indicates that students’ self-regulated learning and academic confidence are substantially enhanced when combined with structured support and feedback—tools that students without formal scaffolds often lack (Kleimola et al., 2025; Nyman, 2024). These dynamics were apparent in the initial survey responses, which reflected ambivalence toward textbook comprehension and anxiety about content mastery prior to lecture.

Post-course responses revealed that the GenAI-generated podcasts helped alleviate some of these barriers. Students described the audio content as a “starting point” that facilitated comprehension and built confidence before engaging with the textbook or attending class. Multiple students cited the podcasts as useful for multitasking or revisiting material prior to exams. This aligns with the cognitive load theory in multimedia learning, which suggests that layered, dual-channel input—such as narrated audio paired with visual or textual information—can facilitate memory encoding and comprehension (Enríquez et al., 2023; Fan et al., 2025; Mayer, 2002; Wakefield et al., 2023).

Quantitative data supported these perceptions. In both courses, students who consistently completed the podcast-linked quizzes—anchored in textbook content rather than podcast transcripts—demonstrated significantly stronger academic performance. Completion rates were positively correlated with higher final course grades, stronger performance on individual exams, improved attendance, and higher scores on notebook-based assessments. These patterns echo prior research on accountability-driven pre-class assignments (Clinton-Lisell, 2026) and suggest that the podcasts may have served as both affective and cognitive scaffolds, easing the transition into more effortful reading tasks (Dunlosky et al., 2013). Importantly, these trends align with findings by (Bassett et al., 2020), who reported that students who prepared using both textbooks and video-based resources performed better in flipped biology courses, and complement broader evidence that active learning approaches significantly improve student performance in STEM education (Freeman et al., 2014). Notably, the strength of the observed correlations—especially between quiz completion and overall course grade (r = 0.8055 in Ecology; r = 0.7793 in Environmental Biology)—indicates that this was not simply a proxy for general academic diligence. Students who used the podcast–quiz system also tended to demonstrate increased engagement across other dimensions of the course. This reinforces the idea that generative AI–supported materials can complement, rather than replace, deeper engagement with traditional academic resources like textbooks and structured note-taking systems.

While correlations between podcast-linked quiz engagement and academic performance were detected across both courses, it is important to further clarify their interpretive scope. These associations should be read as exploratory indicators of a relationship between structured pre-class engagement and student performance, rather than as evidence of a causal effect of the podcasts themselves. Correlational analyses serve an important role in educational research by identifying patterns that warrant further, more targeted investigation, particularly in complex instructional settings where multiple dimensions of student engagement co-occur (Dunlosky et al., 2013; Freeman et al., 2014; Panadero, 2017). In this context, the observed associations point to meaningful links between structured pre-class engagement and student performance, while also raising important questions about how different dimensions of study behavior interact in content-intensive courses.

At the same time, these findings are especially meaningful in the context of ongoing shifts in higher education, where students increasingly struggle with sustained engagement in dense academic reading and often encounter course material for the first time during lectures (McMurtrie, 2024; Pilotti et al., 2023). Identifying consistent associations between pre-class engagement tools and performance outcomes represents an important first step in determining whether such interventions merit deeper empirical attention. Future research could build on this work through matched-cohort or longitudinal designs that control for prior academic preparation, demographic variables, and baseline study behaviors, allowing for more precise examination of instructional sequencing and causal mechanisms (Wang & Fan, 2025). Additionally, while student self-report provided valuable insight into how learners perceived and integrated the podcasts into their preparation process, complementary cohort-based designs would allow this work to advance toward a more programmatic understanding of how generative AI–supported scaffolds function within diverse student populations (Dunlosky et al., 2013; Panadero, 2017).

In-class dynamics shifted noticeably during the intervention semester. Lecture discussions reflected greater familiarity with key terms and concepts, allowing the instructor to move beyond definitional content toward higher-order synthesis and application. These shifts echo long-standing observations that active learning—even in non-majors biology courses—can foster more student engagement and deepen conceptual understanding (McClanahan & McClanahan, 2002). Office hour attendance increased, particularly around exam periods, with students requesting clarification on content they had already previewed through the audio–quiz workflow. This behavior contrasts with patterns in prior semesters, where students often waited until after poor assessments to seek help. While attendance remained steady overall (approximately 85%), the audio scaffold appeared to help occasional absentees maintain pace with the course. These shifts suggest not only improved preparation but also a broader cultural change in students’ approach to learning—moving from reactive to proactive engagement (Tanner, 2012). This pattern mirrors findings from STEM-focused active learning programs like Emerging Scholars, where collaborative and scaffolded environments have been shown to improve both persistence and academic outcomes, especially for underrepresented students (Miller et al., 2021).

Importantly, this intervention demonstrated that GenAI tools such as NotebookLM can be deployed in a pedagogically intentional manner, with instructor oversight guiding content development. Rather than relying on student-initiated queries into open-ended tools such as ChatGPT, the podcasts were scripted using curated chapter material and structured around course objectives. This mitigated risks of AI hallucination and ensured alignment with instructor expectations—a key concern in AI-assisted pedagogy (Dhuliawala et al., 2023; Kasneci et al., 2023). The instructor-controlled nature of the tool also preserved curricular coherence and avoided over-reliance on proprietary publisher platforms.

However, this model also highlights emerging tensions in course design. As publishers increasingly embed AI-generated audio, annotation, and assessment tools into digital textbooks, faculty may face pressure to adopt integrated platforms that offer seamless functionality but limited customization (Selwyn, 2019). While such platforms promise personalization and accessibility, they can erode instructional autonomy, particularly in interdisciplinary or field-based courses that deviate from standardized sequencing. The language of corporate EdTech has drawn criticism for deliberately obfuscating pedagogical aims, with “learnings” emerging as a prime example of how substance is replaced by branding (Nicholls, 2025). These concerns align with critiques by (Kerssens & Van Dijck, 2022), who warn that platformization—particularly through integrated EdTech systems—poses a significant threat to both institutional and professional pedagogical autonomy. Such systems may constrain educators’ ability to design and adapt learning environments in ways that reflect their disciplinary expertise and pedagogical intent.

From a diversity and inclusion standpoint, the use of multimodal GenAI tools may offer culturally responsive benefits, especially for students from historically underrepresented backgrounds who may struggle with academic self-efficacy or lack familiarity with academic norms (Güven et al., 2025; Meyer et al., 2014; Nixon et al., 2024). Audio-based scaffolds provide an accessible, flexible entry point into content and may support students navigating college-level STEM expectations for the first time (Holmes, 2020; Kraal et al., 2022). Global case studies—such as those from El Salvador—highlight that equitable access and ethical frameworks are crucial for ensuring GenAI tools support student learning rather than deepen existing disparities (Valdivieso & González, 2025).

Future work should disaggregate outcomes by demographic characteristics, including first-generation status, multilingual learners, and students with limited prior exposure to disciplinary vocabulary, to assess the equity impact of such interventions (Whitcomb et al., 2021). This study also raises questions about sustainability and instructional labor. While the NotebookLM tool allowed for relatively efficient podcast generation, faculty adoption at scale will require institutional support, including training in GenAI usage and guidance on copyright, data privacy, and accessibility (Symeou et al., 2025). Collaborative approaches between instructional designers and content experts may be needed to ensure high-quality, inclusive implementation across disciplines (Katsamakas et al., 2024).

Future research should examine how podcast length, structure, and tone affect learning outcomes across disciplines, especially for students with limited time or competing responsibilities (Desmedt et al., 2025). Additional inquiry could explore how podcasts align with evolving definitions of educational media, considering factors like production quality, audience interaction, and integration with other tools (Rime et al., 2022). There is also a need to evaluate how GenAl supports adaptive content delivery, automated feedback, and inclusive assessment, while maintaining transparency and ethical alignment (Denny et al., 2024). Recent studies on AI-driven tutoring systems show that such tools can significantly enhance learning progress in distance and hybrid environments, offering promising models for integration in higher education (Möller et al., 2024).

At the institutional level, future studies could assess how the adoption of AI tools influences systems of governance, faculty workload, and access to educational technology (Katsamakas et al., 2024). In STEM education, future research should explore these effects in lower-division and general education biology courses, where foundational reading skills are still forming (Enríquez et al., 2023; Wakefield et al., 2023). Longitudinal studies could evaluate whether repeated exposure to GenAI-enhanced pre-class scaffolds improves reading fluency, content retention, and persistence in STEM. Finally, comparative research could assess the benefits and drawbacks of instructor-generated versus commercially produced AI supports. As the educational landscape continues to evolve, grounded and evidence-based integrations of GenAI will be essential for ensuring that technology supports—not supplants—the deeper cognitive practices that underlie scientific learning.

5. Conclusions

This study examined the role of GenAI audio podcasts, coupled with weekly quiz assessments, as a scaffold for textbook engagement in upper-division biology courses. In doing so, it responded to a pressing pedagogical challenge: students are increasingly arriving to class without having engaged with assigned readings, a pattern exacerbated by the disruptions of the COVID-19 pandemic. Traditional reading expectations no longer align with the media habits and learning environments many students now inhabit. Instructors are witnessing firsthand the consequences of this shift—students encountering course material for the first time during lecture, unable to build upon prior exposure or arrive prepared to participate meaningfully.

Against this backdrop, our intervention piloted a low-cost, scalable GenAI strategy to reintroduce pre-class preparation in a form that resonates with students’ current modalities of engagement. The GenAI-generated podcasts were never intended to replace reading. Rather, they offered a point of reentry—a scaffold that made the textbook’s structure and vocabulary less daunting, helped students begin the process of knowledge acquisition before class, and increased the likelihood of deeper engagement during class. Many students expressed surprise at how much more they understood during lecture simply by having listened to the podcast and completed the quiz beforehand. For some, this created a new sense of ownership over their learning; for others, it was the first time they felt confident entering a complex biology classroom discussion.

Our findings demonstrate that structured, intentional GenAI interventions can enhance academic performance when embedded into weekly learning routines. Students who engaged consistently with the podcast–quiz system exhibited stronger exam preparedness and improved course performance overall. These outcomes suggest that GenAI tools, when implemented as part of an intentional pedagogical strategy, can support both motivational and cognitive components of learning. Importantly, these tools offer instructors an opportunity to meet students where they are—not by lowering expectations, but by designing scaffolds that help students rise to meet them.

This exploratory intervention offers a model for other postsecondary science instructors seeking to adapt their teaching to a post-pandemic educational landscape. It affirms the value of rethinking not just what we teach, but how we invite students into the learning process. While we cannot yet conclude whether this intervention increased textbook reading in the long term, we find strong evidence that it improved pre-class preparation, student confidence, and learning outcomes. Future research might explore how GenAI scaffolds could be integrated with reading-specific strategies—such as guided annotation tools or academic support services—to further cultivate the kind of deep, sustained engagement with scientific texts that remains vital for success in STEM.

As institutions continue to grapple with shifting student behaviors, faculty have an opportunity—and perhaps an obligation—to explore creative solutions that preserve the intellectual rigor of higher education while acknowledging its new realities. The GenAI podcast intervention presented here represents one such step: a way to help students show up more prepared, more confident, and ultimately more empowered in their learning journeys.