Teachers’ Ecological Transformation in Artificial Intelligence Literacy: A Case Study on the Transition from an Anthropocentric to an Ecocentric Perspective

Abstract

The aim of this study is to determine teachers’ views on integrating sustainable artificial intelligence use into classroom teaching processes. The study was conducted using a qualitative research approach and adopted a case study design. The study group consisted of 38 teachers who were selected using maximum diversity sampling, who currently use AI, and who participated in a 4-week structured “Sustainable AI Training Program.” To ensure methodological triangulation, data were collected through semi-structured interviews, researcher diaries, and participant diaries and analyzed using inductive thematic content analysis. According to the analysis results, some findings reveal that teachers considered filtering AI tools through a pedagogical filter centered around the question “Is it really necessary?” rather than using them directly and intensively. Furthermore, digital minimalism was adopted in classroom practices, along with the use of a single, optimized prompt instead of trial-and-error queries, the practice of archiving and reusing generated content, and a shift toward low-tech alternatives. It was determined that teachers would adopt digital minimalism in classroom practices, aiming to serve as role models for sustainable use by bringing the hidden environmental costs of technology into the learning process and fostering eco-digital citizenship awareness among students. Consequently, AI integration has evolved from a technical decision into a pedagogical redesign process encompassing ethical and ecological dimensions.

1. Introduction

The integration of artificial intelligence (AI) and digital technologies into educational processes offers significant opportunities in terms of personalized learning opportunities and pedagogical efficiency [1]. However, current educational technology discourses predominantly focus on human-centered (anthropocentric) pedagogical benefits that optimize learning outcomes [2,3,4,5,6], leaving the enormous ecological costs associated with the production, operation, and use of these technologies out of the discussion [7,8,9]. To put it in concrete terms, a standard Google search requires an average of 0.3 Wh of energy, while a ChatGPT query consumes approximately 2.9 Wh of energy [10]. Training a single large language model (LLM) can produce 284 tons of CO₂, which is equivalent to five times the emissions emitted by an average car over its lifetime [9,11,12]. The increasing use of AI at this level of intensity in training creates the risk of “digital pollution” as an invisible form of digitalization and seriously contradicts the United Nations Sustainable Development Goals (SDGs), particularly the “Responsible Consumption” (SDG 12) and “Climate Action” (SDG 13) [13,14].

Despite these ecological risks, teacher candidates and practicing teachers mostly view AI solely as a functional and pedagogical tool, and they are not sufficiently aware of the cost arising from the hardware, water, and energy consumption behind the technology [11]. This situation, which can be conceptualized as “ecological blindness” in the literature, stems largely from the fact that existing professional development programs for teachers focused on technology integration concentrate heavily on technical skills [11]. However, it is critically important for teachers to develop a critical awareness that enables them to question not only “how” AI tools are used when incorporating them into the classroom environment, but also “at what cost” they are used [15]. This digital minimalism adopted in micro-level classroom practices carries a profound pedagogical value rather than being a quantitative solution that will reverse global carbon emissions on its own [16]. Teachers’ conscious choices to limit technology are thought to create a powerful role model mechanism for raising eco-digital citizens by demonstrating an ethical stance towards technological consumption to students. In this context, there is a great need for evidence-based studies that redefine the role of AI in education within the sustainability paradigm and focus on teacher practices [17,18]. There are studies in the literature that examine teachers’ perceptions and experiences regarding the integration of artificial intelligence into classroom practices. For instance, Alavizadeh et al. (2022) observed that educators generally regard AI tools as beneficial resources for improving instructional efficiency, personalization, and content creation, while also voicing concerns about pedagogical control and ethical ramifications [19]. Similarly, in their systematic review of AI use in K–12 education, Crompton and Burke (2023) noted that existing research primarily focuses on learning outcomes, performance improvement, and instructional effectiveness, showing limited interest in broader socio-ethical dimensions [20]. Additionally, studies by Le et al. (2026) [18] and Ding et al. (2024) [1] indicate that teacher professional development programs tend to address AI literacy primarily in terms of technical proficiency and instructional integration skills. However, these studies largely exclude teachers’ classroom practices and perceptions regarding the ecological dimension of AI use—specifically issues related to energy consumption, carbon footprint, and environmental sustainability. Recent studies on the use of large language models (LLMs) in the context of sustainable education have shown that while these technologies have the potential to transform learning processes, this transformation carries risks without pedagogical guidance and structured use. Indeed, the study conducted within this scope reveals that the use of LLMs in learning processes based on sustainable development enhances students’ critical thinking, independent work, and problem-solving skills, while emphasizing that this effect is largely dependent on structured and guided usage conditions [21]. The same study notes that unsupervised use, however, increases the risks of superficial learning, the production of misinformation, and cognitive laziness. These findings indicate that the use of artificial intelligence in sustainable education is not merely a matter of technological integration; rather, it must be addressed in conjunction with pedagogical design, teacher guidance, and structured usage strategies. Global policy frameworks have also emphasized the need to reconsider the relationship between education, digital transformation, and ecological sustainability. The report “Reimagining Our Futures Together: A New Social Contract for Education” by UNESCO (2021) conceptualizes the contemporary world as a set of interconnected crises, including climate change, biodiversity loss, unsustainable resource use, and rapid technological transformation [22]. These overlapping challenges highlight that education can no longer be structured solely around efficiency, innovation, or economic productivity; rather, it must be reoriented toward sustainability, collective responsibility, and the common welfare The report further underscores the need to rethink curricula, redefine the role of teachers as transformative agents, and develop educational practices that respond to both ecological and digital challenges simultaneously. From this perspective, the integration of artificial intelligence into educational processes should not be evaluated only in terms of pedagogical effectiveness or technological advancement. Instead, it must be aligned with broader ecological and ethical considerations that address the environmental consequences of digital technologies. Therefore, examining teachers’ practices through the lens of sustainable AI use becomes critical for constructing an education system that responds to both the climate crisis and digital transformation. Building on this gap, this study aims to reveal teachers’ views on integrating sustainable AI use into classroom teaching processes.

2. Theoretical Framework

2.1. Transition from AI Literacy in Education to Sustainable Artificial Intelligence

Traditional technology literacy and technology acceptance models (TAM) define teachers’ fundamental barriers as a lack of technical knowledge and the adaptation process [7,19,23,24,25]. However, AI literacy cannot be limited to merely using tools effectively; it also requires critically evaluating the socio-technical relationships, political preferences, and ecological costs behind these tools [26]. It is essential for teachers to focus on the concrete limitations of “actually existing AI” rather than speculative “hype” narratives [27,28].

In this context, the “Green AI” approach, which centers on energy efficiency and reducing the carbon footprint, offers a more sustainable alternative to the “Red AI” paradigm, which prioritizes model accuracy and performance at all costs in educational settings [29,30,31,32]. Rather than remaining dependent on constantly cloud-based systems in the use of educational technologies, the adoption of the principles of “digital degrowth” and “frugal computing” has become a fundamental requirement for sustainable education [33,34,35]. In this context, it is emphasized that for LLMs to be effectively utilized in sustainable educational processes, it is necessary to develop not only technological capabilities but also structured usage frameworks. Indeed, a recent study noted that the effectiveness of LLM use depends largely on “prompt quality” and structured guidance processes; it was found that guided use enhances students’ critical thinking skills and deepens the learning process [21]. This situation demonstrates that a sustainable AI approach must not be limited to energy efficiency or carbon footprint considerations alone, but must also incorporate pedagogical structuring processes.

2.2. The Tension Between Anthropocentric and Ecocentric Approaches in AI Ethics

Current ethical approaches to AI use are generally limited to anthropocentric (human-centered) perspectives that focus on human benefit, such as data privacy, algorithmic bias, transparency, and fairness [36,37]. In contrast, the ecocentric (nature-centered) perspective, which focuses on elements such as the extraction of raw materials required for hardware production, the generation of electronic waste, and energy consumption that threatens planetary boundaries, is largely invisible in educational technology discourse [34].

In the “Anthropocene” era, it is important to acknowledge that digital technologies—including artificial intelligence—can cause indirect environmental harm beyond the benefits intended for humans. This phenomenon is conceptualized as “ecological colonialism,” referring to the exploitation of natural resources, energy consumption, and ecological degradation that disproportionately affect non-human beings and ecosystems. From this perspective, ecological colonialism should be addressed as a pedagogical and ethical issue in education, just as significant as widely discussed topics such as algorithmic discrimination [38,39]. The “transparency” of AI systems must be addressed not only in terms of how algorithmic decisions are made (black box) [40], but also in terms of the reality of “what these algorithms cost” the ecosystem. This holistic approach requires teachers to evaluate technology not only through pedagogical advantages but also through the lens of “eco-digital citizenship,” which relates it to the planet’s ecological boundaries [41,42]. In other words, teachers’ awareness of ecological colonialism at the macro level should be directly reflected in their lesson plans, material choices, and sustainability communication with their students at the micro level. However, the shift from an anthropocentric to an ecocentric perspective should be understood not as a simple shift in ethical priorities but as a multilayered process of pedagogical transformation. Within an anthropocentric framework, artificial intelligence is primarily evaluated in terms of its contribution to human learning efficiency, performance, and instructional effectiveness [39]. An ecocentric perspective expands this evaluative framework by integrating the ecological ramifications of technological utilization, including energy consumption, resource extraction, and environmental degradation [38]. This shift entails a fundamental restructuring of pedagogical reasoning, marking a shift from the question “How can AI improve learning outcomes?” to “Is the use of AI justified given its ecological cost and necessity?” [43]. In this sense, this shift is not merely about incorporating environmental concerns alongside existing criteria; rather, it is transformative, for it redefines what constitutes “responsible” and “effective” pedagogy. Therefore, the transition from anthropocentric to ecocentric can be conceptualized as a shift from technology integration focused on efficiency toward a responsibility-oriented pedagogical stance where teachers critically balance pedagogical benefits with ecological sustainability.

2.3. Teacher Agency and Changes in Pedagogical Intentions in Classroom Practices

Teacher agency refers to teachers’ capacity to make pedagogical decisions based on their own beliefs, values, and goals [44]. In the context of sustainable AI integration, this agency implies a critical stance that involves viewing technology not as a neutral tool but as something that can be limited in its use when necessary or consciously withdrawn (technological ascesis).

Teachers’ awareness of the hidden costs of AI tools on non-human beings and the planet triggers an “ethical awakening” in them, leading to a re-examination of classroom practices [27,45]. This transformation process enables teachers to move away from being “technicians” who passively apply technology and become “change agents” who demonstrate a values-based pedagogical stance [46]. Consequently, ecological awareness, supported through professional development programs, should become an integral ethical dimension of teachers’ pedagogical decision-making processes and be reflected in teaching strategies that offer students an eco-digital citizenship perspective. Examining how teacher agency, combined with this ecological awareness, translates into action is of critical importance for the Green AI literature in education. In this context, the aim of the study is to deeply examine teachers’ views on integrating sustainable AI use into classroom teaching processes.

3. Materials and Methods

This study aims to deeply examine teachers’ views on integrating sustainable AI use into classroom teaching processes. To this end, a qualitative research approach was adopted, and a case study design was used, which allows for a multidimensional analysis of the phenomenon within its own real-life context [47,48]. The case study consists of the experiences of teachers who have undergone a structured training process on sustainable AI use and plan to transfer this awareness to their own classroom environments.

Figure 1 illustrates the overall research process, including participant selection, the four-week structured Sustainable AI Training Program, data collection through methodological triangulation, and inductive thematic analysis. This schematic representation provides a holistic overview of how the study progressed from intervention to interpretation.

3.1. Participants

The study group consists of 38 teachers working in state schools affiliated with the Ministry of National Education in Turkey during the 2025–2026 academic year. Participants were selected using maximum diversity sampling, one of the purposive sampling methods, to provide rich data appropriate to the focus of the study [49,50]. To implement the maximum diversity sampling strategy, an initial pool of approximately 62 teachers who had participated in the Sustainable Artificial Intelligence Training Program organized by researchers and affiliated institutions was identified. From this pool, participants were purposefully selected to ensure diversity across key characteristics such as teaching fields (preschool, elementary school, science, Turkish, and social studies), years of professional experience, and gender. The selection process aimed to capture a broad range of perspectives rather than focusing solely on statistical representativeness. Teachers meeting the inclusion criteria were first filtered, and then a balanced distribution across the defined diversity dimensions was ensured. As a result, 38 teachers representing maximum diversity across these characteristics were included in the study. Participants were selected through an invitation process conducted in collaboration with school administrations affiliated with the Ministry of National Education. First, teachers who met the study’s participation criteria were invited to participate. Participation in the study was entirely voluntary. Participants were purposefully selected based on predefined criteria to obtain information-rich cases. The main selection criteria were that teachers had previously received training in AI use and had actively used AI-supported applications in their classrooms for at least one year. Overall, 22 participants were female, and 16 were male, ranging in age from 23 to 38. The distribution by subject area was as follows: preschool (n = 12), classroom teaching (n = 9), science (n = 7), Turkish language (n = 5), and social studies (n = 5). In terms of professional experience, 17 teachers had 1–3 years of experience, 12 had 3–5 years, and 9 had 5 years or more. The research was conducted on a voluntary basis, the necessary ethical approvals were obtained, and participants were anonymized and coded as (T1, T2 … T38). The research process strictly adhered to various ethical principles, in addition to obtaining approval from the ethics committee. Participation in the study was entirely voluntary, and informed consent was obtained from all participants prior to data collection. Participants were clearly informed about the purpose of the study, the procedures to be followed, and their right to withdraw from it at any stage without facing any adverse consequences. To protect privacy and confidentiality, no personally identifiable information was gathered, and all data was coded and kept safe. Additionally, audio recordings and written data were used solely for research purposes and were not shared with third parties. During the reporting process, care was taken to present findings in a manner that prevents the identification of participants and protects their professional and personal integrity. The implementation process is outlined in

Table 1. Implementation process.

Week	Focus	Theoretical Component	Practice-Based Activities	Data Generated
Week 1	AI Literacy and Ecological Awareness	Lifecycle of AI systems (development, deployment, usage); energy consumption, carbon footprint; Green AI vs. Red AI concepts	Analysis of sample AI-supported classroom scenarios; group discussions on potential environmental impacts; reflection writing	Baseline awareness reflections; notes on existing AI classroom practices
Week 2	Environmental Impact of Classroom AI Use	Hidden environmental costs of digital technologies (cloud processing, storage, repeated queries); sustainability in digital pedagogy	Development of an Environmental Impact Matrix analyzing participants’ own AI-supported teaching practices	Environmental impact matrices prepared by participants
Week 3	Sustainable Pedagogical Integration of AI	Sustainability-oriented TPACK perspective; responsible and necessity-based AI use; efficiency-focused instructional planning	Redesign of an existing lesson plan considering ecological AI use; peer feedback sessions	Sustainable lesson plan drafts
Week 4	Development of Sustainable Classroom Strategies	Student awareness of ethical and ecological AI use; integrating sustainability into classroom learning environments	Design of mini-projects or classroom activities promoting sustainable AI awareness; final reflective reports	Strategy reports and activity/project designs

.

Teachers’ opinions were gathered not through a purely theoretical discussion, but following a four-week educational intervention specifically designed for the study. The training program was designed and implemented directly by the researchers. Throughout the four-week process, the researchers actively facilitated theoretical sessions, guided practice-based activities, and monitored participants’ reflections through their diaries. This direct involvement enabled the researchers to ensure consistency in content delivery and to closely observe participants’ developmental processes. This program consisted of the following stages: the life cycle and ecological costs of AI systems (Red vs. Green AI) in the first week, the hidden environmental costs of digital technologies in the second week, the restructuring of lesson plans with sustainability-focused TPACK integration in the third week, and the design of sustainable AI awareness projects for students in the fourth week. This structured context prepared the ground for teachers to translate their sustainable pedagogical decision-making processes into action.

3.2. Data Collection Tools and Data Collection Process

In this study, methodological data triangulation was used to reveal teachers’ thoughts on how to integrate the sustainable use of AI into classroom teaching processes in a multidimensional way. Data triangulation is one of the fundamental validity strategies that strengthens the consistency of research results [49]. Semi-structured interviews and researcher and participant diaries were used together to increase the credibility of the research and the consistency of the findings.

Semi-Structured Interviews: In order to thoroughly determine teachers’ ethical awareness regarding AI use, their pedagogical decision-making processes, and their sustainability-focused classroom approaches, researchers developed a semi-structured interview form. The content of the form was reviewed by four academics specializing in educational sciences, teacher education, educational technologies, and assessment for its scope and research purpose, and finalized. The semi-structured interview form is presented in Appendix A. The use of a semi-structured interview form was preferred because it allows participants to express their experiences within their own contexts and enables the in-depth and comprehensive identification of perception, interpretation, and implementation processes related to the phenomenon under investigation [48,49]. During the interviews, probing questions such as “Can you explain your thoughts on this with an example?” were asked as appropriate to help teachers concretize their experiences and pedagogical decisions within their own contexts. In addition to the semi-structured interview form as a data collection tool, researcher diaries were used to observe teachers’ classroom practices throughout the training process. In addition, all semi-structured interviews were conducted following completion of the four-week educational intervention program. Consequently, the interviews allowed participants to reflect in depth on their experiences, the transformations they underwent, and their pedagogical decision-making processes following the structured training.

Researcher Diaries: Going beyond the structured views expressed in the interviews, researcher diaries were used to directly observe teachers’ reflections throughout the training on sustainable AI use. These diaries were used to reflect participants’ experiences and awareness development throughout the process.

Participant Diaries: Participant diaries were used to reflect the internal transformation and awareness developed by teachers regarding AI technologies and their ecological costs during the four-week training process from their own subjective perspectives. These diaries, containing individual reflections, enriched the data set by making visible the cognitive reasoning processes that cannot be directly observed and the pedagogical restructuring processes over time.

The diaries are an important data source that captures the participants’ evolving thoughts over time and makes visible the cognitive and ethical reasoning processes that the researcher cannot directly observe [51]. In this respect, the diaries contributed to tracking teacher agency and professional reflection regarding the sustainable use of artificial intelligence. Data triangulation, achieved by using multiple data sources in the study, aimed to increase the credibility and verifiability of the findings [52,53]. Data triangulation is one of the fundamental validity strategies that strengthens the consistency of research results by comparing information obtained from different methods and data types [49]. Thanks to this approach, teachers’ views, application behaviors, and individual reflections were analyzed together, developing a multi-layered understanding of sustainable AI use. In this study, data collection took place after the training program. This is because the primary objective of the study was not merely to describe teachers’ current status, but rather to conduct an in-depth examination of the processes of awareness, transformation, and pedagogical restructuring that emerged following the sustainable artificial intelligence training program.

3.3. Data Analysis

The multiple data sets obtained in this study (interview transcripts, researcher and participant diaries) were analyzed using an inductive thematic content analysis approach, in which patterns of meaning were extracted directly from the data without adhering to predetermined categories [48,54]. Face-to-face interviews with teachers lasted approximately 30–35 min, and all interviews were audio recorded and transcribed. Following the interviews, participants were asked to listen to their recordings again and confirm their statements, applying member checking. This practice is an important credibility strategy used to confirm the accuracy of participant statements and increase the extent to which the interpretations reflect participant experiences [50,53]. The inclusion of direct quotations from teacher statements in the findings section also strengthened the traceability of the interpretations based on the data. During the analysis process, all interview transcripts, researcher and participant diaries were read repeatedly to develop a comprehensive understanding of the research context. Subsequently, the texts were examined line by line, meaningful statements related to sustainable AI use were converted into open codes, and a data-driven coding approach was adopted [55]. MAXQDA 2020 qualitative data analysis software was used in the coding and data management process. In this study, MAXQDA 2020 was used as a systematic data organization, coding, and access environment to support the inductive analysis process. First, all interview transcripts and diary entries were imported into the software as separate document groups. Each data source was organized within MAXQDA to enable comparisons across data types. During the initial coding phase, meaningful units were identified through line-by-line reading, and open codes were assigned directly within the software. These codes were created inductively and improved over time by comparing them to each other. MAXQDA’s coding system enabled researchers to visually cluster similar codes and progressively develop higher-level categories and themes. While the software’s “Code System” and “Document System” functions were used to track relationships between codes and data sources, the “Selected Sections” function allowed researchers to review all excerpts associated with a specific code and ensure consistency in interpretation. Additionally, researchers examined code frequencies and distributions across different data sources using MAXQDA’s built-in analytical tools, which supported the identification of dominant patterns. Furthermore, MAXQDA facilitated the comparison of independently coded data segments during the inter-coder reliability process. Coded segments were reviewed within the software environment, and disagreements among coders were discussed by directly examining the associated data excerpts. This process provided transparency and traceability regarding how themes were derived from the raw data. After coding with the MAXQDA 2020 program, a comprehensive coding framework containing frequency and percentage values was created, and the distribution between themes was supported by quantitative indicators. To ensure coding reliability, two researchers independently coded 40% of the data set. The agreement rate between independent codings was calculated as 85%. In qualitative research, a concordance level of 80% or above is considered an acceptable indicator of reliability [54]. Therefore, the remaining data set was coded by one researcher. Discordant codes were discussed among the researchers and reorganized based on consensus, and the agreed-upon codes and themes were used in the final analysis. This process was conducted using a controlled analysis approach that strengthened the dependability of the research [53]. To increase the reliability of the findings, the interview data were compared with the researcher and participant diaries, thus cross-validating the data sources. In addition, some of the findings were read back to the participants to obtain feedback on the accuracy and clarity of the interpretations. This feedback, obtained through participant validation, was used to assess whether the researcher’s interpretations reflected the participants’ actual experiences [50]. In the later stages of the analysis process, it was observed that no new codes emerged, existing codes were repeated, and thematic patterns stabilized. The repetition of similar themes indicates a decrease in new information production and the attainment of data saturation. Data saturation is one of the fundamental criteria used in qualitative research to decide when to terminate the data collection process [56]. Accordingly, it was concluded that the themes obtained in the study were sufficient to represent the phenomenon.

4. Results

The findings obtained from the content analysis are presented in five thematic subsections. Each subsection focuses on a specific dimension of teachers’ experiences and decision-making processes regarding sustainable AI use, including pedagogical transformations, decision criteria, environmental awareness, instructional planning, and recommendations.

4.1. Changes in Teachers’ Pedagogical Approaches to AI Use

This subsection presents the changes in teachers’ pedagogical approaches to AI use following the training program, focusing on their shift from task-oriented to purpose-driven practices.

Table 2. Teachers’ views on changes in their approaches to the use of AI in the classroom.

Theme	Category	Code	f
Transition from task-oriented to goal-oriented	Re-establishment of the reason for use	“Is it necessary?” questioning	33
		“Why am I using it?” questioning	32
		Removing AI as the first source of reference	30
Redefining the concept of efficiency	Expanding the scope of efficiency	Holistic decision-making based on pedagogical benefit, ethical responsibility, and ecological awareness	33
		Responsible production instead of rapid production	30
Translating ecological cost awareness into decisions	Invisible cost information	Energy–water consumption	30
		Carbon footprint awareness	20
		E-waste	18
	Integrating environmental information into ethics	Viewing the ecological dimension as part of the ethical field	28
Query optimization and reduction of digital consumption	Prompt strategies	Writing targeted prompts in one go	31
		Reducing multiple small queries with a single comprehensive query	26
		Summary and targeted output instead of long output	24
		Creating templates and reusable content	20
	Critique of digital consumption	Awareness of digital consumption and consumption culture	19
Classroom organization in pedagogical practice	Prioritizing the student’s thought process	Brainstorming before AI	25
	Collaborative production	Producing and editing output together as a class	23
	Limiting usage	Planned use with limitations within the lesson	20
The role of the teacher	Professional positioning	Transition from consumer to conscious regulator role	19
	Guidance role	Instilling a culture of minimal but effective use in students	18
Integrating the environmental dimension of AI into teaching content	Integration into the lesson	Class discussions on the relationship between sustainability and technology	16
	Discipline-specific application	Calculating the environmental cost in lesson outcomes	13
	Adaptation to the student’s level	Simplifying to a child’s level	12
Shifting towards alternative methods instead of AI	Low-cost alternatives	Peer assessment	16
		Manual editing	14
		Open-source material	13

shows that nearly all of the 38 participants adopted the question “Is it necessary?” as a pedagogical filter in the use of artificial intelligence. This high frequency proves that the awareness created by the training module among teachers is not accidental, but rather indicates a strong trend toward “purpose-driven transformation” shared across the group. Participants stated that they began to question the necessity of its use in line with pedagogical goals, rather than positioning AI as a tool used directly and intensively. For example, classroom teacher T4 stated, “…in the processes within the classroom, I will shift from the question ‘how do I use AI tools?’ to ‘why should I use them?” They stated that this change brought about a more conscious and selective approach in instructional decision-making processes…”. Science teacher T9 said, “Before the training, I evaluated AI tools more in terms of pedagogical efficiency, speed, and content production capacity. I actively used AI in processes such as preparing experiment handouts, generating alternative questions, adapting activities to different levels, and analyzing student responses, especially in science lessons. However, in my practice of using these tools, environmental aspects such as energy consumption, water usage in data centers, carbon footprint, or hardware e-waste production were almost completely absent. I was thinking about AI literacy in terms of technical, ethical, pedagogical, and critical dimensions, but the “ecological cost” remained an invisible area in this framework…”. The change in teachers’ understanding of efficiency was also reflected in the participants’ and researchers’ diaries. For example, in T12’s daily, “…I used to think that getting an activity plan from ChatGPT in seconds was a professional achievement. Today, when I consider the water and energy costs behind this action, I now question whether this plan is really necessary…” In the researcher diaries, the emotional change experienced by teachers while discussing the concepts of speed and efficiency was clearly observed. In the note at the end of the second week of the researcher diaries, “…I observed in the participants’ (especially T9 and T12) expressions that their initial proud attitude of ‘the fastest user of technology is the best’ gave way to deep astonishment and a slight sense of guilt after seeing the energy cost data. Ö12’s question, “While I thought I was doing something useful, was I actually harming the planet?” symbolized the moment when the AI in the classroom evolved into a cautious responsibility…” Teachers’ confrontation with invisible ecological costs triggered a deep professional reckoning within them. The astonishment and slight guilt experienced by T9 and T12, who had viewed their ability to quickly use AI before training as a source of ‘professional success’ and ‘pride,’ after being confronted with the energy data, is the most obvious sign of this internal reckoning. T12’s reaction, “I thought I was doing something useful, but was I actually harming the planet?”, is a striking example of how the ‘ecological blindness’ [11] discussed in the literature is broken in teacher practice. This moment of breaking represents a moment of ‘triggered ethical awakening’ [27,45], when the teacher stops seeing AI as a neutral tool that only provides benefits and realizes that their actions have planetary consequences. It has been determined that the understanding of efficiency is not limited to speed and time savings but is redefined by the joint assessment of ethical responsibility, learning quality, and environmental impacts. Awareness of the invisible ecological costs of AI systems, such as energy consumption, water usage, and carbon footprint, has increased. In line with this awareness, it has been understood that teachers tend to reduce unnecessary digital consumption by optimizing queries, creating shorter and more purpose-driven requests, and using selected content that supports the learning process instead of long outputs. Social studies teacher T13 stated, “…In the past, when students asked questions, I would directly obtain content from AI and project it onto the screen. Now, I first brainstorm with the students, then complete the missing points with a shorter and more targeted prompt…” These statements show that AI is being repositioned as an auxiliary element that supports students’ thinking processes rather than a tool that directly produces content in the teaching process. It is seen that pedagogical balance is being sought through practices such as brainstorming before class, creating joint outputs, and limiting the duration and scope of use. In this process, it was determined that the role of the teacher has also transformed from a technology consumer to a guide who regulates the learning process, and that instilling a culture of “less but effective use” in students is considered part of their professional responsibility. Some participants stated that they would begin to integrate the relationship between sustainability and technology into their course content, include discussions on environmental costs in discipline-based activities, and, in some cases, turn to low-cost alternative methods such as peer assessment, manual editing, and the use of open-source materials instead of artificial intelligence. Overall, the findings reveal that the training process did not merely increase teachers’ technical competence but also fostered a more critical and sustainable professional awareness that integrates pedagogical, ethical, and ecological responsibilities related to AI use.

4.2. Criteria Guiding Teachers’ AI Use Decisions

This subsection examines the criteria that teachers consider when deciding whether and how to use AI tools in their classroom practices.

Table 3. Teachers’ opinions on the criteria they consider when deciding to use AI tools in their lessons.

Theme	Category	Code	f
Decision-Making with Pedagogical Justification	Pedagogical necessity	Pedagogical value-added test	28
	Pedagogical purpose alignment	Clarity of learning outcomes and objectives	25
	Deepening learning	Avoiding superficiality	23
Ecological Sustainability and Awareness of ‘Invisible Costs’	Environmental cost awareness	Sensitivity to energy, water, and carbon footprint	18
	Minimum usage	Using only what is necessary and consciously	16
	Limiting high-density production	Reducing production requiring intensive processing	12
Establishing the Efficiency–Sustainability Balance	Time–cost balance	Time savings alone are not a sufficient justification	15
	Minimalism	Maximum pedagogical output with minimal processing	14
Query Management	Prompt quality	Single and optimized prompt	28
	Reducing repetition	Not reproducing the same output	20
Prioritizing Alternatives and Technological Minimalism	Low-cost alternative	Alternative method instead of AI	14
	Collaborative and face-to-face learning	Focusing on student production	10
Ethical Responsibility and Data Governance	Expanding the ethical framework	Ethical and ecological integrity	13
The Teacher’s Role as a Role Model	Serving as a model	Providing students with a sustainable usage model	17
	Raising awareness	Translating energy costs into classroom language	14
Institutional and Long-Term Sustainability Perspective	Collective sustainability	Sharing and common materials	10

shows that teachers will now adopt a multidimensional evaluation process when deciding to use AI tools in their lessons. Participants primarily consider pedagogical justification as the fundamental criterion, stating that they question the use of AI in terms of “contributing to real learning” and “alignment with learning objectives,” and consciously avoid uses that only increase speed but do not create depth of learning. For example, classroom teacher T10 described this situation as follows: “…I used to routinely use tools like ChatGPT when writing activity plans. However, now, if I can produce the same plan myself in a short time or use existing material with a minor revision, I will not opt for AI use. Because every query means a processing load and therefore energy consumption…” Similarly, preschool teacher T1 stated, “…Before training, my main criterion for using AI was ‘time savings’. Now, I will try to strike a balance between time savings and environmental costs. If a use saves me 5 minutes but requires significant processing power, I will reevaluate that use…” Science teacher T9 also stated, “… If I am going to use AI to generate hypotheses in the process of designing experiments with students, I will evaluate whether this deepens the learning process. Producing something that requires high processing power just to save time or to produce ready-made content no longer seems like a sufficient pedagogical justification to me…” However, it was observed that environmental sustainability was included in the decision-making process, that unnecessary and computationally intensive production would be limited in line with awareness of invisible costs such as energy consumption and carbon footprint, and that the principle of minimal and necessary use was adopted. Participants do not consider time savings alone to be a sufficient justification, they have stated that they will strive to strike a balance between efficiency and sustainability, aiming to produce fewer but pedagogically high-quality outputs. In this context, it has been stated that query management and optimization strategies have been developed, and that digital consumption will be reduced by writing single, well-structured queries, reducing repetitive production, and limiting output volume. Social studies teacher T2 stated, “…If an activity can be done without using AI, I will prioritize methods such as face-to-face discussion, collaborative learning, or primary source analysis…” Similarly, mathematics teacher T6 stated, “…Instead of generating 50 different questions for a math skill, I will prefer to generate 8–10 high-quality questions appropriate for the teaching objective and discuss them in depth. Instead of constantly requesting re-production, I will use a single, optimized prompt…” It was understood that teachers would also prioritize alternative pedagogical methods over artificial intelligence, thus moving towards an understanding of technological minimalism. Awareness has also developed regarding ethical responsibility and data management, data privacy, and minimum data sharing, and sustainability criteria in platform selection have been considered. In addition, it has been observed that teachers position themselves as guides offering a sustainable usage model to students, aiming to foster digital citizenship awareness by bringing topics such as energy costs and conscious technology use into classroom discussions. Finally, some participants pointed to a collective sustainability approach that goes beyond the individual level, questioning school/institutional policies and aiming to reduce reproduction through shared materials. Turkish teacher T16 stated, “…When we get help from artificial intelligence, could there be an invisible energy cost?” I think this discussion will contribute to raising conscious individuals who question technology. When using AI tools in my lessons, I will now pay attention to the environmental dimension in addition to the ethical issues I previously considered…” These findings reveal that decisions regarding the use of AI have evolved from being a technical choice to a comprehensive evaluation practice that integrates pedagogical, ethical, and ecological responsibilities.

4.3. Impact of Environmental Cost Awareness on Decision-Making

This subsection explores how teachers’ awareness of the environmental costs of AI influences their classroom decision-making processes.

Table 4. Teachers’ views on the impact of awareness of the environmental costs of AI on decisions regarding its use in the classroom.

Theme	Category	Code	f
Change in decision-making logic	Reframing decision criteria	Pedagogical necessity filter	33
		Incorporating ecological cost into decision-making	31
Frequency of use and production economy	Reorganizing frequency of use	Conscious reduction	29
	Productivity management	Collective–planned use	28
		Avoiding reproduction	25
		Prompt optimization	25
	Resource utilization strategy	Material reuse	23
Pedagogical repositioning in usage	Changing the role of tools	Shifting to a perspective that supports thinking from the product producer towards AI tools	27
	Alternative production methods	Using existing resources	25
	Learner-centeredness	Prioritizing student production	20
	Classroom organization	Creating joint output	19
	Low digital intensity	Turning to physical alternatives	15
Transformation reflected on students and classroom culture	Awareness teaching	Digital sustainability awareness among students	17
	Role modeling	Redefining the teacher as a model	12

shows that awareness of the environmental costs of AI technologies, such as energy and water consumption, will lead to a significant reframing of teachers’ decisions regarding classroom use. Participants stated that they now evaluate AI use through a pedagogical necessity filter based on the question “Is it really necessary?” and that information about environmental costs will become a criterion included in the decision-making process. Science teacher T19 stated, “…After gaining awareness of the environmental costs of AI technologies, such as energy and water consumption, I can say that I experienced a clear mental shift in my classroom usage decisions. Previously, I mostly evaluated AI tools in terms of pedagogical efficiency, time savings, and content diversity. However, after this training, knowing that every use creates energy consumption and an indirect carbon footprint in the background added a new ethical and ecological dimension to my decision-making process…” Social studies teacher T16 also stated, “…Before becoming aware of this, I used AI tools quite frequently, especially in processes such as lesson planning, question generation, and creating instant explanations. We even produced short content together with students in class. But now, when using AI tools, I will ask myself this question: Is this use truly pedagogically necessary, or am I choosing it simply because it is convenient? I will decide whether to use it or not based on the answer to this question…” stated. During the research process, teachers’ acquisition of ecological awareness was not a linear and smooth process; a distinct emotional transition phase was experienced. As reflected in the researchers’ diaries (Week 3), participants such as T2 and T16 initially perceived the usage restrictions as a threat to their professional efficiency and showed a tendency to protect their habits, stating, “All these details discourage us from using technology.” However, this resistance broke down when they experienced the quality output of a single optimized prompt, giving way to a feeling of “relief”. This situation shows that the ecological awakening among teachers occurred not only through the transfer of theoretical knowledge, but also through the experience that digital minimalism is more pedagogically satisfying. In this direction, a tendency to strike a balance between pedagogical benefit and ecological burden has emerged. The use of AI has been repositioned as a decision area involving ethical and environmental responsibility, rather than being a technical or practical preference. It has been determined that increased awareness has led not to a complete abandonment of usage frequency, but to a conscious reduction. Teachers have developed collective and planned usage strategies instead of fragmented and momentary queries, avoided repetition, and optimized processes with fewer but higher-quality requests. Approaches based on production economics, such as reusing and adapting produced content, evaluating existing materials, and turning to manual preparation or low-digital-intensity alternatives, have also attracted attention. For example, mathematics teacher T23 stated, “…Whereas I used to produce separate problem sets for 8th grade learning outcomes, I now create a multi-level question pool with a single optimized prompt and use it over a long period of time. This reduces the number of interactions with the system. But I don’t intend to stop using it completely. My only goal is to succeed in acting from a perspective of efficiency…” The role of AI in the classroom has also been redefined, shifting from a tool that directly produces content to a model where it is used as an auxiliary tool that supports the teacher’s thinking. It has been stated that applications that prioritize student production, group-based work, and physical activities will be preferred more. In addition, teachers stated that they aim to develop digital and ecological sustainability awareness in students, that they will bring the environmental impacts of AI use into classroom discussions, and that they will position themselves as role models who demonstrate conscious usage behavior. For example, preschool teacher T26 said, “… I was producing outputs from AI tools for many activities. However, I will now start to refocus on child-centered physical materials. We will now design cards with the children using recycled materials. I will carry out activities related to the correct use of technology, and I plan to mention the possible harms of using AI tools in order to raise awareness…” These findings reveal that environmental cost awareness will transform AI usage into a pedagogically justified, planned, and sustainable usage culture rather than reducing it.

4.4. Planned Instructional Adjustments for Sustainable AI Use

This subsection presents the instructional adjustments teachers plan to implement in order to ensure sustainable AI use in their lesson planning.

Table 5. Teachers’ views on the arrangements they plan to ensure sustainable AI usage in their lesson plans.

Theme	Category	Code	f
Transforming AI into a justified pedagogical choice	Need-based decision making	Is it necessary? filter	27
	Making visible at the lesson plan level	Adding a field for justification of AI usage	25
Reducing unnecessary and repetitive use	Reducing frequency of use	Reducing unnecessary and habit-based usage	25
	Batch and planned use	Batch and scheduled queries	24
	Reuse and archiving	Long-term and multi-purpose use of generated output	20
	Prompt quality	Maximum output with minimal querying	20
	Limiting high processing power	Limiting production that requires intensive processing	14
	Setting behavioral boundaries	Weekly usage quota	13
Pedagogical sustainability	Returning to low technology	Shifting towards alternative pedagogical methods	10
	Leaving production responsibility to the student	Student production first, AI second	10
	Redefining the role of AI	Positioning AI as a feedback and final control tool	9
Integrating the ecological-ethical dimension into lesson plans and transforming student awareness	Developing student awareness	Teaching ecological awareness and environmental ethics	8
	Content integration	Linking to sustainability gains	8
	Collaborating on the decision-making process	Discussing the decision to use AI with students	7

shows that teachers plan to make conscious and systematic adjustments to their lesson plans in order to use AI tools more sustainably after the training. Participants stated that they adopted a necessity-based decision-making approach, evaluating the use of AI before each activity by asking, “Is it really necessary?” They also indicated that they considered adding explanatory sections to their lesson plans that highlight the pedagogical contribution, environmental cost, and alternatives of using artificial intelligence. In this process, practices aimed at reducing unnecessary and habit-based use, developing bulk and planned usage strategies instead of instant and fragmented queries, preventing re-production by reusing generated content, and ensuring output optimization with fewer but higher-quality prompts have come to the fore. For example, classroom teacher T15 stated, “…Previously, when preparing weekly plans, I would generate separate content for each lesson. Now, I plan to use bulk and optimized prompts to get structured output in one go. I will also use existing textbooks and materials from my own archive instead of AI for simple learning outcomes. I will also ask students to use AI for specific tasks rather than for every assignment…” Furthermore, the limitation of productions requiring high processing power, the conscious imposition of quotas on frequency of use, and the reflection of the digital minimalism approach—which prioritizes measured technology use—in the plans have been noteworthy. Within the scope of pedagogical sustainability, teachers showed a tendency to turn to alternative teaching methods, focus on student production, and position AI as an auxiliary tool that provides feedback and editing support rather than a tool that directly produces content. Furthermore, integrating ecological and ethical dimensions into course content, developing environmental awareness among students, designing activities linked to sustainability outcomes, and discussing AI usage decisions with students to foster a shared culture of responsibility were among the planned applications. Another classroom teacher participating in the research, T16, stated, “…I realized that I needed to model the sustainability dimension of education not only at the content level but also at the application level. When explaining environmental responsibility to students, it would be inconsistent for me to remain dependent on high-energy-consuming tools in my own teaching practices. Therefore, I now plan to have students write their own stories first in story-writing activities and use AI only as a feedback tool. Additionally, in projects requiring visual production, instead of creating separate visuals for each student, we will produce a shared example and analyze it as a class. I will also increase the use of physical materials in activities like STEM to reduce the need for digital production…” These findings reveal that teachers are moving towards a planned and sustainable teaching design that integrates pedagogical purpose, resource efficiency, and environmental responsibility, rather than a superficial limitation on AI use.

4.5. Teachers’ Recommendations for Sustainable AI Integration

This subsection outlines teachers’ recommendations for integrating AI into classroom practices in a sustainable and environmentally responsible manner.

Table 6. Teachers’ recommendations for sustainable AI use that considers environmental impacts.

Theme	Category	Code	f
Conscious use in the balance of pedagogical necessity and environmental cost	Necessity filter	Use only when there is a genuine pedagogical need	35
	Proportionality	Remove AI from being the center of the lesson and position it as a supporting component.	32
	Temporal arrangement	Use it weekly and with planning	28
Reducing workload and efficient production	Query efficiency	Reduce retries with structured prompts	34
	Reducing repetition	Existing material archive	32
		Progress through textbooks and use AI to improve and meet specific needs.	30
	Shared production	Create a shared pool within the school.	29
	Output management	Reduce storage	25
	Limiting high-processing production	Avoid unnecessary visuals and high resolution	25
Classroom sustainability pedagogy and digital citizenship	Pedagogical diversity	Preserve and use low-energy alternative methods	30
	Student awareness	Raise awareness of ecological AI	25
Institutionalization and program integration	School policy creation	Develop corporate policy and guidelines	21
	AI-L framework in teacher training	Integration of the ecological dimension into AI	21
Deepening ethical stance	Prudence	Question its necessity	20

shows that participants’ suggestions for the use of sustainable AI that takes environmental impacts into account are concentrated in the dimensions of pedagogical necessity, digital minimalism, classroom sustainability pedagogy, institutional regulation, and ethical awareness. Teachers emphasized that AI should not be used automatically as a solution tool, but should only be employed when there is a genuine pedagogical need, and should be positioned as a supporting component rather than the centerpiece of the lesson. In this regard, it was recommended that its use be planned temporally, that purposeful production be carried out in a single session, that trial-and-error-based repetitions be reduced, that existing materials be improved and reused, and that duplicate production be prevented by creating shared content pools. Participants also noted the need to limit visually intensive productions requiring high processing power, avoid unnecessary data storage, and adopt practices that reduce digital consumption. In classroom applications, it was recommended to preserve low-energy alternative teaching methods (such as discussion, drama, drawing, and nature-based learning), to develop awareness among students about the invisible energy and carbon costs of artificial intelligence, and to address conscious usage behavior as a pedagogical goal. In addition, it was stated that sustainable AI use should not be limited to individual efforts, that guiding principles should be developed at the school level, that the ecological responsibility dimension should be systematically integrated into teacher training on AI literacy, and that institutional application consistency should be ensured. Another point emphasized by the participants was the need to approach technological innovations critically and cautiously, questioning the necessity and alternatives before each use. For example, classroom teacher T1 stated, “…When I learned that cloud storage also consumes energy, I realized that the habit of accumulating unnecessary files should be abandoned…” Similarly, social studies teacher T17 stated, “…The savings of a single teacher may be limited. However, if a conscious usage policy is established throughout the school, the impact will be greater…” Science teacher T2 stated, “…Teachers and students should acquire effective prompt writing skills. Furthermore, organized usage is more sustainable than unplanned and intensive individual usage…” … And Turkish teacher T4 stated, “…Digital tools should not eliminate pedagogical diversity. Furthermore, students should not only be users but also conscious digital citizens…” In the final week of the training, the researcher’s diaries indicate that the pressure on teachers to keep up with technology has decreased and been replaced by professional confidence. For example, “… Teachers now seem to have overcome the anxiety of being passive consumers of technology. In particular, the excitement on T26’s face as he described his plan to combine AI with recycled materials in the classroom shows that sustainability is perceived as a ‘creative space’. As a researcher, I have concretely observed that teachers develop suggestions in a collective solidarity and trust, saying to each other, ‘We don’t have to do this, we can use less energy this way’…” These findings reveal that the sustainable use of AI requires a holistic change supported not only by technical regulations but also by pedagogical choices, institutional policies, and ethical awareness.

5. Discussion

This study was conducted to determine teachers’ views on integrating sustainable AI use into classroom teaching processes. For this purpose, a Sustainable AI Education Program was developed. Teachers from different disciplines participated in this training program. During this training process, researchers and participants kept diaries. At the end of the training, data was collected through semi-structured interviews with the participating teachers. These data were described through content analysis.

5.1. Changes in Teachers’ Pedagogical Approaches to AI Use

The findings of this study show that there has been a clear paradigm shift in teachers’ pedagogical approaches to the use of AI. The shift of participants from the question “how do I use it?” to “why should I use it?” after the training aligns with the human-centered and critical AI approach emphasized in the literature in recent years. Holmes and colleagues (2022) state that AI in education should be positioned not as a productivity tool independent of pedagogical purpose, but as a system that empowers human agency and supports instructional decisions [57]. Similarly, Selwyn (2022, 2023) argues that using generative AI tools in education based on speed and automation logic can lead to pedagogical reductionism, and therefore a critical filter is needed [27,58]. In this study, teachers’ repositioning of AI not as a direct content-producing authority but as an auxiliary tool supporting students’ thinking processes shows a strong parallel with the aforementioned theoretical frameworks. Another finding that emerged in the research is the redefinition of the concept of efficiency. While participants evaluated AI in terms of speed, time savings, and content production capacity before the training, after the training, they developed a holistic approach that also took into account ethical responsibility, learning quality, and environmental impact. This change is consistent with the assessments of Kasneci and colleagues (2023) regarding the opportunities and risks that large language models offer in education [59]. The authors note that LLMs can only produce pedagogical benefits with conscious limitation and guidance. Crompton and Burke’s (2023) systematic review of AI research at the K-12 level also reveals that applications are mostly evaluated in terms of learning outcomes and performance improvement, while the sustainability dimension is largely neglected [60]. In this context, the current study is considered to fill an important gap in the literature by addressing the concept of efficiency alongside ecological responsibility. Teachers’ awareness of the hidden costs of AI systems, such as energy consumption, data center water usage, and carbon footprint, can also be linked to discussions on the environmental impact of machine learning. The carbon footprint of large language models presented by researchers such as Strubell et al. (2019), Henderson et al. (2020), Patterson et al. (2021), and Luccioni et al. (2022) has highlighted the carbon footprint of large language models and the Green AI literature, which has largely remained within the domain of computer science and macro-level policymakers to date [12,61,62,63]. The most significant novelty this research contributes to the field is that it takes the ecological cost of AI (water consumption, energy expenditure, and carbon footprint) out of the realm of abstract technological and engineering discussions and transforms it into a ‘concrete pedagogical micro-design variable’ that teachers take into account in their lesson plans, material choices, and classroom practices. Indeed, in our research, teachers moving beyond the mere motivation of saving time to engage in query optimization, creating single, targeted prompts to avoid redundant production, is the actual and actionable counterpart of these macro-scale global energy debates in classroom pedagogical practices. Teachers’ shift towards query optimization, creating shorter and targeted prompts, and avoiding unnecessary output production demonstrates that macro-level energy debates find their counterpart in the pedagogical sphere. The findings also point to a change in the role of teachers. UNESCO’s (2023) guide [57] on generative AI and the OECD’s (2023) [64] policy notes particularly emphasize the teacher’s role as a critical mediator and ethical guide. In this study, teachers positioning themselves as guides who regulate the learning process rather than as technology consumers is consistent with these international policy documents. However, the current research expands the framework by adding the dimension of ecological responsibility to this guiding role. Participants’ plans to integrate sustainability discussions into their course content and, in some cases, to prefer peer assessment or traditional methods over AI, demonstrate the development of a culture of conscious and balanced use, rather than technology opposition. In line with Selwyn’s (2023) critique of “techno-solutionism,” it can be said that a critical awareness has emerged regarding the avoidance of technology-dependent pedagogical approaches [58]. Ultimately, these findings show that AI literacy should not be limited to technical and ethical dimensions; the ecological cost dimension must also be integrated into teacher training programs. The study demonstrates that sustainable digital pedagogy is possible at the level of classroom practices and emphasizes that AI use requires a higher level of ethical and ecological sensitivity. In this respect, the research deepens the discussions in the educational technology literature on responsible and sustainable AI use through concrete teacher practices.

5.2. Criteria Guiding Teachers’ AI Use Decisions

The direct reflection of the macro-level global energy crisis on micro-level classroom practices is one of the most significant contributions this study offers to the literature. Indeed, teachers’ decisions to use AI tools have evolved from being a choice based solely on technical efficiency or time savings to a multi-layered evaluation practice that integrates pedagogical, ethical, and ecological dimensions. In particular, teachers’ statements such as “every query means a processing load and energy consumption” directly correspond to findings regarding the high energy consumption of large language models and deep learning systems. Strubell, Ganesh, and McCallum (2019) [12] and Patterson et al. (2021) [62] have shown that the training and use of large-scale language models generate significant carbon emissions, while Henderson et al. (2020) [61] argue that the carbon footprint of machine learning studies should be systematically reported. Calculations performed by Luccioni, Viguier, and Ligozat (2022) [63] on the BLOOM model have also made the concrete environmental costs of production processes visible. In this context, teachers’ decisions to avoid unnecessary and computationally intensive productions demonstrate that the sustainable AI approach discussed at the macro level in the literature is reflected in micro decision-making processes within the classroom. Van Wynsberghe’s (2021) [9] “Sustainable AI” framework emphasizes that AI must be sustainable not only for sustainability but also in terms of its own production and use processes. These findings show that the theoretical framework corresponds to the pedagogical practice level. Teachers’ refusal to view time savings as a sufficient justification on its own and their questioning of AI use based on its real contribution to learning and alignment with objectives parallels the critical literature on educational technologies. Selwyn (2022) [27] argues that digital technologies in education should be evaluated based on pedagogical value rather than narratives of inevitable progress; he proposes a critical stance against technological solutionism. Similarly, Macgilchrist, Allert, and Bruch (2020) [65] and Selwyn (2024) [34] note that low ecological impact and “degrowth” approaches are gaining increasing importance in scenarios regarding the future of educational technologies. UNESCO’s (2023) [57] guidance document on generative AI also emphasizes that AI use must be based on clear pedagogical justification and be consistent with ethical and sustainability principles. In this context, teachers’ preference for conducting activities that can be done without AI through face-to-face discussion and collaborative learning reflects an approach that positions technology not as a substitute for pedagogy, but as a tool serving pedagogical goals. This indicates a conscious departure from the superficial and demonstrative forms of use criticized in the literature as “AI theatre” [66]. The findings show that teachers develop strategies such as single and optimized requests, reducing repetition, and producing fewer but higher-quality outputs. This approach is consistent with discussions of digital minimalism and digital degrowth. In an educational context, digital degrowth advocates for the adoption of meaningful and measured use rather than increasing technological intensity [27,67]. In this study, teachers’ preference for focusing on fewer, high-quality questions rather than producing many questions indicates a pedagogical–ecological shift from quantity to quality. Thus, conscious limitation of digital production is adopted instead of its unlimited expansion, and a balance is established between pedagogical depth and environmental responsibility. Findings in the ethical and data management dimensions are also consistent with the existing literature. Bender and colleagues (2021) emphasized that large language models carry not only technical but also social and ethical risks, drawing attention to issues of data usage, transparency, and accountability [68]. UNESCO (2023) [57] has also established data privacy and minimum data sharing principles as the fundamental normative framework for the use of generative artificial intelligence. Teachers positioning themselves as role models for sustainable and conscious technology use among students demonstrates that AI use intersects with digital citizenship education. In this respect, the findings reveal that AI use is not only a matter of instructional design but also a matter of value transfer and social responsibility. Finally, the tendency of some teachers to shift their decision-making processes from the individual level to the institutional policy level shows that sustainability is understood as a collective responsibility. Hou (2025), who discusses the typology of AI integration in education, states that the policy and governance framework at the institutional level is decisive [69]. Similarly, the findings of this study show that the sustainable use of AI should not be limited to individual awareness; it must be supported by collective strategies such as sharing common materials, reducing reproduction, and reviewing school policies. Overall, these findings reveal that AI usage decisions have evolved into a new form of professional reasoning that integrates pedagogical competence, ethical responsibility, and environmental sustainability dimensions. This indicates that AI integration in education must be rethought around more conscious, justified, and sustainable technology use. Thus, teachers are positioned not merely as practitioners who use technology, but as responsible actors who reflect the ecological and ethical boundaries of the digital age in their pedagogical decisions.

5.3. Impact of Environmental Cost Awareness on Decision-Making

The research findings show that awareness of the environmental costs of AI tools creates a normative reframing in teachers’ pedagogical decision-making processes. This finding is consistent with the current literature highlighting the impact of AI on increasing global energy consumption and carbon footprint. For example, Pimenow, Pimenowa, and Prus (2024) note that the increasing computational power requirements of AI systems directly impact global energy consumption and create a new risk area in the context of climate change [70]. Similarly, Dere (2026) emphasizes that AI’s contribution to carbon emissions is often overlooked in policy and implementation discussions, defining this area as a neglected source of emissions [71]. Wang, Li, and Li (2024) empirically examine the impact of AI on the ecological footprint and energy transitions, drawing attention to the complex relationship between digitalization and environmental pressures [72]. These macro-level discussions are concretized at the micro-pedagogical level by making the question “Are teachers really necessary?” a decision filter in your work. The tendency toward optimization rather than reduction that emerges in your findings is also consistent with the energy-aware AI literature. Różycki, Solarska, and Waligóra (2025) define the development of energy-aware machine learning models as a fundamental requirement for sustainable artificial intelligence [73]. Hakimi, Tarashtwal, and Ghafory (2025) state that the Green AI approach should be based on the principles of less computation, less repetition, and more conscious model use [74]. In our study, teachers creating long-term content pools with single, optimized prompts, avoiding repetitive production, and developing planned usage strategies can be read as the practical counterpart of this theoretical framework in an educational context. Furthermore, studies highlighting the energy and water consumption aspects of AI also support your findings. Morain, Ilangovan, and Delhom (2024) systematically examine the effects of AI applications on water consumption and resource management, revealing that digital technologies should be evaluated not only in terms of energy but also in terms of their water footprint [75]. Ogundiran, Asadi, and Gameiro da Silva (2024) systematically address the effects of AI on energy efficiency and environmental quality, emphasizing that digital technologies must be considered alongside sustainability goals [76]. In the context of educational structures, Tariq et al. (2024) examine the use of complex AI models for energy sustainability in educational buildings, demonstrating that the field of education cannot be considered independently of energy debates [77]. When evaluated in light of this literature, the current study points to three important aspects. The findings show that the sustainable integration of AI in education is shaped not only at the technical infrastructure or policy level, but also in teachers’ micro decision-making processes. They indicate that sustainable AI pedagogy can be built through teacher awareness and ethical self-regulation.

5.4. Planned Instructional Adjustments for Sustainable AI Use

According to the research results, rather than superficially limiting the use of AI after training, teachers tend toward conscious, planned, and reasoned pedagogical restructuring. This situation parallels the pedagogical redesign approach proposed by Toha (2025) within the framework of sustainable higher education [78]. The fact that the teachers in the study ask the question “Is it really necessary?” before each activity is a concrete manifestation of the goal-oriented integration principle emphasized in the literature [79] at the classroom level. Furthermore, this questioning strongly aligns with the literature on digital minimalism, digital degrowth, and sustainable educational technologies. In particular, Selwyn’s (2024) digital degrowth approach emphasizes the need to adopt pedagogically and ecologically defensible uses of technology in education rather than its automatic expansion [33]. In this study, teachers developed planned, optimized, and collective usage strategies instead of habit-based and fragmented inquiries, which conceptually parallels Selwyn’s proposed radical sustainable educational technology perspective. Teachers’ use of collective and optimized prompts, reducing reproduction, and developing content reuse strategies demonstrates micro-level awareness towards reducing energy consumption and carbon footprint. This finding aligns with the recommendations of Shengjergji et al. (2024) and Zawacki-Richter and Cefa (2024), who highlight the invisible environmental costs of EdTech and suggest that carbon calculations should be made visible in an educational context [80,81]. Participants’ plans to limit high-computational-power productions, reduce reproductions, and optimize outputs relate to discussions on energy intensity and carbon footprint emphasized in the sustainable AI literature. Rapha (2024) states that technical, economic, and ecological dimensions must be considered together in the context of sustainable artificial intelligence, particularly noting the increased energy consumption of large-scale models [82]. De Chirico et al. (2024) highlight the invisibility of energy costs in the digital realm and recommend adopting minimalist and adaptive strategies [83]. Teachers’ approach of setting usage quotas and generating fewer but higher-quality requests aligns with the energy-sensitive design principles proposed in this literature. Findings indicate that teachers are beginning to treat environmental cost not as an abstract concept but as a concrete design variable at the planning level. In the context of pedagogical sustainability, teachers positioning AI as an auxiliary tool providing feedback and editing support rather than a direct content-generating tool aligns with responsibly orchestrated mediation approaches. Bakar and Tapsoba (2026) propose a “responsibly orchestrated mediation” framework for the use of AI in the classroom, emphasizing the importance of teachers maintaining pedagogical control [84]. Similarly, Wu, Zeng, and Song (2025) concluded that generative AI plays a supportive role in design education, but pedagogical goals must remain central [85]. The findings of this research show that teachers’ conscious choices to preserve pedagogical autonomy and center student production are consistent with the balancing models proposed in the literature. Integrating ecological and ethical dimensions into course content and discussing AI usage decisions with students is also related to the perspective of sustainability literacy. Ott (2024) argues that sustainability communication in the digital age requires not only content transfer but also the development of critical digital awareness [86]. Ambar and Sena (2025) show that the principles of energy efficiency and minimalism can be integrated into instructional design in AI-supported digital education designs [87]. In this context, the current findings reveal that teachers have begun to address sustainability not only at the subject level but also at the pedagogical modeling level. The emphasis on modeling in the findings reinforces the principle of pedagogical coherence, which is frequently mentioned in the literature. Teachers’ plans to balance the need for digital production by increasing the use of physical materials are parallel to studies that question the assumption that digitization is automatically sustainable. Baxter (2025) highlights the invisible environmental costs of cloud technologies, showing that digital solutions often result in high energy consumption based on material infrastructure [88]. Teachers in this study also appear to have developed a critical awareness that digital production is not always more environmentally friendly. This aligns with the call to make invisible costs visible, emphasized in discussions on sustainable AI [82]. Based on these findings, awareness training on environmental and invisible costs generally leads to a rethinking of the epistemological and ethical foundations of instructional design rather than a superficial technology restriction. When the pedagogical contribution of AI tools is evaluated alongside energy consumption and alternatives, teachers position technology use not as an automatic indicator of innovation, but as an area of decision-making that requires responsibility. This situation highlights the need for structured approaches to sustainable AI literacy in teacher education programs.

5.5. Teachers’ Recommendations for Sustainable AI Integration

Sustainable AI use necessitates an institutional and systematic transformation beyond individual ethical preferences. The study found that teachers position technology integration not merely as a technical issue but as a holistic transformation area within the context of pedagogical justification, digital minimalism, institutional policy development, and ethical responsibility. This aligns significantly with the current literature on how AI is addressed in teacher education. Daher’s 2025 thematic review of AI in the context of teacher education emphasizes that AI literacy should be approached within an expanded framework that includes critical and ethical dimensions, not merely technical skills [89]. Similarly, Ariza and Mountford (2025) state that AI literacy must be integrated with the principles of sustainability and inclusivity, drawing attention to the importance of keeping pedagogical design central in teacher preparation programs [90]. These findings directly align with the participants’ emphasis that “AI should be a supporting component, not the center of the lesson.” Findings related to environmental cost and invisible energy awareness are also parallel to the sustainable AI literature. De Chirico and colleagues (2024) advocate for the development of design strategies for energy sustainability in the digital realm, specifically recommending the critical evaluation of applications requiring high processing power [83]. Participants’ suggestions to limit visually intensive productions requiring high processing power and avoid unnecessary cloud storage are consistent with these studies. These findings show the reader that teachers now view environmental sustainability not as a macro-level policy debate but as part of micro-pedagogical decisions within the classroom. The dimensions of digital minimalism and conscious use are also reflected in the literature. Ott (2024) argues that a new literacy paradigm is needed in the context of sustainability communication in the digital age and that the critical, measured, and purposeful use of digital tools has become a pedagogical necessity [86]. Ambar and Sena (2025) suggest integrating energy efficiency and minimalism into instructional design in sustainability-focused educational design [87]. Participants’ suggestions, such as purposeful production in a single session, reducing trial-and-error repetitions, and creating shared content pools, demonstrate that digital minimalism can be transformed into a pedagogical strategy. This situation reveals that the use of sustainable AI must evolve from individual awareness to institutional collaboration. From the perspective of classroom sustainability pedagogy, the findings show that the environmental impacts of AI can be made a direct teaching goal. Capecchi et al. (2025) reveal that AI-supported applications can be used to develop sustainable consumption and digital literacy [91], while Wu, Zeng, and Song (2025) state that generative AI can be integrated into pedagogical models in the context of sustainable design education [85]. Osman (2026) emphasizes that AI literacy should be linked to sustainable development goals [92]. The participants’ views on educating students not only as users but as conscious digital citizens are consistent with this literature. This finding shows that the use of sustainable AI is a pedagogical goal that encompasses not only teacher behavior but also the development of students’ values and attitudes. Finally, the emphasis on the need for institutional regulation and policy, when evaluated together with participant views pointing to the limitations of individual action, aligns with calls for systemic transformation in the literature. Mahajan (2025) emphasizes the importance of structural regulations when discussing the effects of AI on cognitive and social sustainability [93]. Carrillo et al. (2025) state that teacher education programs need to be restructured within an institutional framework [94]. In this context, the findings clearly show that the sustainable use of AI cannot be limited to individual ethical preferences; it must be supported by school policies, teacher education curricula, and institutional guiding principles. Overall, based on these findings, it offers a unique contribution that is consistent with the current literature, which extends AI literacy from technical skills to the dimensions of ecological and ethical responsibility, and concretizes it through teacher experiences.

6. Conclusions

This study was conducted to examine teachers’ views on integrating the sustainable use of artificial intelligence (AI) into classroom teaching processes, particularly in the context of the tension between human-centered efficiency and environment-centered sustainability. The findings clearly demonstrate that the research objective was achieved and that the study successfully addressed its central research question. First, the results indicate that teachers have undergone a significant shift in their pedagogical reasoning regarding AI use. Participants have begun to adopt a critical decision-making framework focused on the question “Is this approach really necessary?” rather than merely focusing on efficiency, speed, and content production. This shift demonstrates that the educational intervention effectively fostered a deeper awareness of the ethical and ecological dimensions of AI use. Second, the study reveals that awareness of ecological costs—particularly regarding energy consumption, water usage, and carbon footprint—has become an integral part of teachers’ instructional decision-making processes. This finding directly addresses the research objective of examining how sustainable AI perspectives can be integrated into classroom practices. Third, teachers’ practices have evolved from habitual and intensive use toward a more planned, optimized, and minimal use of AI tools. Strategies such as command optimization, reuse of generated content, and prioritizing low-tech alternatives signal a transition toward sustainable digital pedagogy. Overall, the findings indicate that the integration of AI in education should be evaluated not merely as a technical or efficiency-focused process but as a multidimensional pedagogical design issue that encompasses ethical responsibility and ecological sustainability. This study contributes to the literature by demonstrating that sustainable AI use can be implemented at the micro-level of classroom practices through teachers’ awareness and initiative. In conclusion, the study confirms that raising awareness of AI’s environmental costs can trigger a meaningful transformation in teachers’ pedagogical orientations. However, sustaining this transformation requires long-term support through institutional policies, teacher education programs, and continuous professional development initiatives.

7. Limitations of the Study

This study aimed to reveal teachers’ views, intentions, and pedagogical planning regarding the sustainable use of artificial intelligence. The dataset relies heavily on teachers’ self-reports. Although the data was triangulated with researcher observation reports, the actual energy savings, water footprint, or reduction in digital carbon emissions generated in the classroom by teachers’ digital minimalism or query optimization strategies were not quantitatively measured. Furthermore, the research implementation process is limited to a four-week structured training and reflection program. Whether participants’ behavioral intentions, such as using pedagogical filters or turning to alternative low-tech methods instead of AI, are sustainable in the long term is beyond the scope of this study. Longitudinal designs are needed to determine whether old habits are resumed over time. These constitute the limitations of this research.

8. Recommendations

Based on the research findings, the following recommendations have been developed for teachers, school administrators, teacher training programs, and future research in order to ensure sustainable AI integration in classroom teaching processes:

Recommendations for Teachers:

• Teachers should use the question “Is using this tool pedagogically necessary, or can the same results be achieved with low-tech alternatives?” as a self-assessment criterion before incorporating AI tools into their lesson plans.
• Single, structured prompt strategies that reduce energy costs should be adopted instead of instant and fragmented in-class queries. Furthermore, unnecessary visual and video productions requiring high processing power should be limited, and any text-based materials produced should be archived for reuse in different contexts.

Recommendations for Institutions and School Administrators:

• The production of repetitive content via AI by different teachers in the school for the same or similar learning outcomes creates a hidden ecological cost. A shared “AI Content Archive” should be established among school departments to reduce unnecessary energy consumption at the institutional level.
• School administrations should develop institutional guiding principles that question not only data privacy but also the energy intensity and sustainability transparency of the AI tools used in decisions regarding digital platform and technology integration.

Recommendations for Policymakers and Teacher Training Programs:

• In current pre-service and in-service teacher training programs, AI literacy is mostly taught in terms of technical skills and academic integrity. Education policies should be updated to make the issue of invisible environmental costs, as indicated by the findings, an integral part of technology integration courses as an eco-digital citizenship and ethics module.

Recommendations for Future Research:

• This study focused on teachers’ statements and plans regarding the sustainable use of AI. Future research should include experimental studies measuring the actual carbon and energy footprint differences between different classes implementing Red AI and Green AI strategies.
• Longitudinal studies examining students’ cognitive awareness levels regarding the environmental costs of AI and the impact of this on their technology usage habits will make significant contributions to the field.