Introduction

Generative artificial intelligence (GenAI) is no longer a hypothetical issue in higher education. It is increasingly embedded in the digital environments students already use to search for information, draft text, and receive automated feedback. In online writing courses, students may turn to tools such as ChatGPT, Gemini, or Microsoft Copilot to clarify assignment prompts, generate ideas, outline essays, summarize readings, or request feedback on developing drafts. At the same time, instructors and institutions continue to debate whether GenAI should be restricted, regulated, or integrated into writing instruction. The Conference on College Composition and Communication (2026) has affirmed the right of students and teachers to refuse generative AI in the college writing classroom, underscoring that disciplinary conversations about these tools remain unsettled. Many institutional responses have emphasized academic integrity and prohibition rather than instructional design. What remains less developed in these conversations is a practical, pedagogical question: how should students be taught to use GenAI as a learning resource?

In online first-year composition, GenAI enters an instructional environment already populated with multiple forms of support, including assignment sheets, rubrics, instructor feedback, peer review, writing tutors, librarians, and learning management system resources. Yet, the availability of support does not guarantee that students will use it strategically. Online learners are often expected to interpret task demands, identify appropriate resources, seek help when needed, and evaluate feedback with limited immediate instructor presence, even though these capacities are not always explicitly taught (Greenhow et al., 2022; Kebritchi et al., 2017). Research on self-regulated learning suggests that effective learners engage in planning, monitoring, and reflection, but these behaviors develop through practice and support rather than access alone (Winne & Azevedo, 2022; Zimmerman, 2000, 2002). Universal Design for Learning similarly emphasizes the importance of flexible supports, learner agency, and executive-function scaffolds that help students navigate tasks, make decisions, and persist strategically (CAST, 2018, 2024; Laist, 2024). Access to options alone is insufficient; students need structured guidance in how to find, evaluate, and use available supports in ways that strengthen self-efficacy, motivation, and metacognitive awareness.

This broader instructional capacity is referred to here as resource literacy: students’ ability to identify, evaluate, select, and strategically use academic supports in ways that advance their learning goals. In online composition courses, this includes interpreting assignment prompts accurately, consulting rubrics and examples, seeking feedback from instructors or peers, using library and tutoring services when needed, and judging whether feedback or assistance is useful for revision. Resource literacy is grounded in motivational and behavioral dimensions of self-regulated learning, including task value, self-efficacy, metacognitive monitoring, effort regulation, and help-seeking (Greene & Azevedo, 2007; Pintrich et al., 1991; Zimmerman, 2000, 2002). Across course iterations in online first-year composition, a recurring pattern emerged: students who struggled were often not students without access to support, but students who did not consistently recognize when help was needed, which resource aligned with the task, or how to evaluate feedback once they received it. GenAI entered this already complex landscape rather than creating it. It made an existing problem more visible.

Emerging research on GenAI in writing instruction reinforces this need for structure, finding that unstructured GenAI use can reduce metacognitive engagement (Bobula, 2024; Deng et al., 2025; Fan et al., 2025), while use paired with reflection, comparison, and critical analysis supports stronger monitoring and more strategic decision-making (Brusilovsky, 2024; Buss et al., 2025; Cheng et al., 2025; Chiu, 2024; Nazari et al., 2021). The central instructional issue, then, is not simply whether students will use GenAI. It is whether online learning environments are designed to help students use GenAI purposefully, transparently, and in relation to other academic supports rather than as a default shortcut.

This article presents CLEAR—Clarify, Locate, Engage, Assess Feedback, and Revise/Reflect—as a practical instructional cycle for structuring how students interact with learning resources across the writing process. Drawing on three years of instructional iteration in online community college composition courses, CLEAR positions GenAI as optional, transparent, and evaluative: one resource among many within a broader ecosystem of academic support. Rather than treating GenAI as a separate problem to solve, the model focuses on helping students make resource decisions that are visible and open to critique. The goal is not simply responsible GenAI use, but stronger resource literacy across the writing process.

Literature Review

The presence of support does not guarantee that students will identify, select, or use resources strategically. In many online contexts, students are expected to interpret task demands, locate appropriate help, evaluate feedback, and revise accordingly, even though these capacities are often assumed rather than explicitly taught. Research on online learning suggests that learners do not always seek help promptly, use available supports effectively, or sustain the motivational and metacognitive habits required for independent academic work (Greenhow et al., 2022; Kebritchi et al., 2017; Rivers et al., 2021; Yot-Domínguez & Marcelo, 2017). Related studies also indicate that many students enter higher education without fully developed self-regulatory capacities for the level of independence college coursework demands (Blackmore et al., 2021; Sáez-Delgado et al., 2023; Uka & Uka, 2020; Vosniadou, 2020).

Research on self-regulated learning consistently emphasizes that planning, monitoring, and reflection are learned behaviors rather than automatic habits. Zimmerman (2002) describes self-regulated learning as a cyclical process in which learners set goals, monitor progress, evaluate outcomes, and adjust future behavior. In online writing environments, these processes become especially important because students often work with less immediate instructor presence and must independently navigate uncertainty, feedback, and revision. Similarly, the Center for Applied Special Technology’s (CAST) Universal Design for Learning framework emphasizes flexible pathways for engagement, representation, and action, but also highlights the importance of executive-function supports that help learners plan, make decisions, and persist strategically (CAST, 2018, 2024). UDL does not assume that access alone creates success; students benefit when instructional design makes strategic decision-making visible and supported. Together, SRL and UDL suggest that effective course design should not only provide resources, but also explicitly teach students how to evaluate and use them.

GenAI has entered an already existing educational landscape, forcing educators to determine the positionality of this new technology within courses. Emerging scholarship positions GenAI as a learning resource that can support idea generation, exploratory thinking, and feedback, particularly during early stages of writing (Hutson & Plate, 2023; Labadze et al., 2023; Phan, 2023). At the same time, researchers caution that unstructured AI use may reduce critical thinking engagement when students accept fluent, plausible-sounding output without sufficient critique or verification (Bobula, 2024; Deng et al., 2025; Fan et al., 2025). In these cases, GenAI may function less as a support for learning and more as a mechanism for cognitive offloading that weakens monitoring and reflection.

By contrast, studies suggest that when GenAI use is paired with reflection and critical analysis, students demonstrate stronger monitoring and more strategic decision-making (Brusilovsky, 2024; Buss et al., 2025; Cheng et al., 2025; Chiu, 2024; Nazari et al., 2021). Requiring students to compare AI-generated feedback with rubric criteria and instructor expectations helps shift GenAI from an answer-producing tool to a prompt for evaluative thinking. This distinction is central to positioning GenAI as an educational resource. The instructional problem is not simply whether students should use GenAI, but whether they are being taught how to evaluate any resource critically, including tools that are immediate and easily mistaken for authoritative support. Faculty conversations, however, often remain centered on policy, ethics, prohibition, and academic integrity rather than on practical routines for embedding GenAI within existing instructional design (Chiu, 2024; Perkins, 2023).

In response to this gap, CLEAR as a repeatable instructional cycle is grounded in five design commitments: GenAI is optional, transparent, evaluated, positioned as one resource among many, and paired with reflection. These commitments align with both self-regulated learning and Universal Design for Learning by preserving learner agency, normalizing help-seeking, supporting metacognitive monitoring, and reinforcing strategic decision-making. Rather than treating GenAI as exceptional, this approach situates it within a broader collection of already existing support. The goal is not to normalize GenAI use for its own sake, but to help students make deliberate choices about when, why, and whether a resource supports their learning. In this way, GenAI becomes less a question of compliance and more a part of resource literacy.

Project Implementation

CLEAR—Clarify, Locate, Engage, Assess Feedback, and Revise/Reflect—was implemented in online first-year composition courses at Cochise College in Southern Arizona across three academic years. The cycle was designed to make resource use visible, intentional, and teachable throughout major writing assignments. Rather than treating GenAI as a separate instructional problem, CLEAR positioned it as one resource among many. This structure also reflects research on GenAI-supported self-regulated learning, which emphasizes that GenAI tools are most useful when they support planning, monitoring, and reflection rather than replace learner judgment (Jin et al., 2023).

Clarify: Teaching Students to Decode the Task

The Clarify phase focused on task interpretation before drafting began. Students completed a short task-analysis activity in a weekly journal designed to interrupt the common tendency to begin writing before assignment expectations were fully understood. They were asked to restate the assignment in their own words, identify the primary rhetorical goal, and list the criteria by which the work would be evaluated. This phase supported the planning dimension of self-regulated learning by making interpretation and goal-setting explicit rather than assumed. In asynchronous writing environments, where students often work with less immediate instructor guidance, this structure helped reduce ambiguity and strengthened early confidence in approaching complex assignments.

At this stage, GenAI was modeled through weekly lecture videos as a questioning tool rather than a drafting tool. Students could prompt GenAI with questions such as, “Based on this assignment description, what questions should I ask to fully understand the task?” They then compared those responses with the assignment sheet and rubric. As part of the journal reflection, students identified at least one place where GenAI misunderstood, oversimplified, or failed to capture assignment expectations. This practice reinforced the principle that GenAI must be evaluated rather than accepted at face value and helped position close reading and critique as the primary work of the phase.

Locate: Resource Mapping Before Drafting

After clarifying goals, students moved into the Locate phase by planning how they would use support before drafting began. In the weekly journal, students identified the most challenging part of the assignment, selected which resources could help address that challenge, and explained why those supports were a good fit. Options included the rubric, assignment sheet, example essays, writing tutors, librarians, peer review, instructor feedback, and GenAI tools. If students selected GenAI, they were also asked to define its intended function, such as brainstorming counterarguments, refining a thesis, or clarifying a concept.

This phase reflected UDL’s emphasis on flexible pathways, executive-function support, and learner agency (CAST, 2018, 2024; Laist, 2024). Students were not simply shown where resources existed; they were required to justify why a particular support aligned with a specific task. This maintained student agency while reinforcing that support selection should be strategic rather than reactive. Since GenAI was optional, students who chose not to use it still moved through the same planning process of determining their primary supports. The instructional focus remained on decision-making, not tool adoption.

When students chose to use GenAI during drafting, they also submitted screenshots of or links to their interactions as part of the journal. This documentation was used for instructional transparency rather than surveillance. It allowed the instructor to see how prompts were framed, how outputs were interpreted, and where students might need additional guidance before questionable use became embedded in final drafts.

Engage: Structured GenAI Use During Drafting

The Engage phase centered on composing in response to the plan established in earlier stages. Students began drafting by acting on their goals and selected supports. For students who chose to use GenAI, the course provided explicit modeling through recorded lecture videos and written examples that contrasted weak prompts, such as “Write my introduction,” with stronger prompts that asked the tool to generate revision ideas, surface counterarguments, or explain possible rhetorical choices. The same modeling also demonstrated how to access non-GenAI supports, including emailing the instructor, accessing tutoring services, librarians, assignment sheets, and rubrics.

This phase was designed to preserve student agency while emphasizing that support tools should extend thinking rather than replace it. When used as an unrestricted efficiency tool, GenAI can encourage cognitive outsourcing and shallow engagement. When structured as one optional support within a larger sequence of writing, it can help students begin, explore, and test ideas while remaining responsible for accuracy and authentic voice. Within CLEAR, GenAI functioned as a drafting stimulus rather than an authoring substitute.

Students who used GenAI during drafting also annotated how they interpreted and revised the output. A typical reflection might explain that an AI-generated topic sentence was revised because it did not connect clearly to the thesis or lacked sufficient specificity. This shifted attention away from what the tool produced and toward what the student chose to keep, revise, or reject. That decision-making process was the instructional target.

Assess Feedback: Comparing GenAI Feedback With Human Feedback

The Assess Feedback phase asked students to compare feedback across sources after a draft had been produced. Students documented feedback received from GenAI, peers, instructors, or other resources and explained which comments were useful or inconsistent with assignment expectations. If GenAI was used for draft review, students were prompted to identify one suggestion that conflicted with the rubric, one area where GenAI feedback was overly general, and one way they verified whether a recommendation was accurate.

This phase supported the monitoring function of self-regulated learning by requiring students to assess the quality of feedback rather than simply receive it. It also reinforced that no resource—human or AI—should be treated as automatically authoritative. The goal was to help students recognize that all feedback must be interpreted in relation to task demands and course criteria, a practice aligned with calls for AI literacy that emphasizes prompt awareness, critical thinking, and human judgment (Tzirides et al., 2024; Walter, 2024). Comparing GenAI feedback against rubric expectations often created what might be described as productive skepticism, helping students become more confident in judging the quality of support rather than deferring to the “fluency” of AI-generated advice.

Revise: Reflection and Strategic Adjustment

The final phase, Revise/Reflect, extended beyond editing a draft. Students reflected on what they learned about their writing process, whether they followed their original goals, which resources they used, and whether those resources helped or hindered progress. In CLEAR, revision was intentionally paired with reflection so that students reconsidered not only the writing they produced, but also the strategies and supports that shaped it. This is aligned with self-regulated learning models that position monitoring and reflection as central to how learners consolidate experience, evaluate strategy use, and adjust future behavior (Winne & Azevedo, 2022; Zimmerman, 2002).

Reflection also helped prevent GenAI from becoming a transactional shortcut. Students often described GenAI as most helpful during brainstorming or early clarification, but less useful for nuanced rhetorical development or final revision. Some noted that the need to verify GenAI outputs made it less efficient than expected, while others recognized that human feedback from peers, tutors, or instructors offered stronger guidance for revision. Students who chose not to use GenAI reflected on how other supports shaped their work, reinforcing that the cycle functioned regardless of tool choice. The purpose of reflection was not to produce uniform responses, but to help students become more deliberate in how they judged, adopted, revised, or rejected available resources.

Across all phases, CLEAR treated strategic resource use as a visible part of the writing process rather than an assumed or hidden one. By embedding planning, support selection, guided engagement, feedback comparison, and reflection into a repeatable sequence, the cycle connected resource literacy, self-regulated learning, and optional human-centered GenAI use within everyday coursework. In this way, GenAI was not isolated as a separate instructional challenge, but incorporated into a broader framework for helping students clarify goals, seek support, evaluate feedback, and revise with greater deliberateness over time.

Project Outcomes

Across three academic years of implementing the CLEAR cycle in fully online first-year composition courses, several recurring patterns emerged in how students used GenAI alongside other academic supports. These observations are course-based observations rather than generalizable findings, but they reveal how structured reflection, transparency, and guided comparison influenced students’ decision-making over time. The most significant shift was not whether students used GenAI, but how they learned to evaluate it in relation to other available resources.

Outcome 1: Students Used GenAI Primarily for Brainstorming and Clarification

Students most often reported in their weekly journals of using GenAI for idea generation, clarification of assignment expectations, early-stage outlining, and feedback on developing drafts rather than for full essay review. Common uses included generating possible counterarguments, refining thesis statements, clarifying assignment language, explaining unfamiliar concepts, and identifying possible starting points for analysis. In this role, GenAI often functioned as a low-stakes entry point into the writing process, particularly for students who struggled to begin drafting or who demonstrated uneven self-regulatory habits, a pattern consistent with studies suggesting that AI-supported writing can increase task engagement, motivation, and writing self-efficacy when used as a structured aid rather than as a replacement for thinking (Hong et al., 2025; Huang & Mizumoto, 2024; Ng et al., 2024; Zare et al., 2025).

At the same time, heavier GenAI use during drafting often produced writing that was polished in tone but weak in specificity. When those drafts were compared against rubric criteria, generic phrasing and misalignment with assignment expectations became easier to identify. Students frequently discovered that fluent language did not necessarily meet rhetorical or disciplinary expectations. In these moments, the rubric functioned not only as an assessment tool, but also as a corrective mechanism that helped students recognize the limits of plausible-sounding but insufficiently tailored writing.

Outcome 2: Reflection and Transparency Reduced Overreliance

Initial overreliance on GenAI appeared to decrease when use was paired with structured reflection in weekly journals and required disclosure through AI transparency statements attached to major essays, which aligns with emerging research emphasizing documentation, reflective use, and human-centered engagement as safeguards against passive or uncritical AI use (Kell et al., 2025; Krajka & Olszak, 2024; Sanders, 2025; Tran, 2024). In earlier iterations, students often documented GenAI use without analyzing whether it actually supported or hindered their thinking. When reflection prompts were expanded, responses became more nuanced. Some students described GenAI as creating additional work because its output had to be verified against sources or revised extensively for specificity and accuracy. Others described it as useful for clarifying ideas while still insufficient for replacing original thought.

Transparency also influenced online classroom dynamics in important ways. When GenAI use was openly discussed through discussion forums rather than treated as a hidden practice, students asked more direct questions about acceptable use, shared examples of AI-generated errors, and appeared more willing to discuss uncertainty. In this context, disclosure functioned as more than an academic integrity measure; it created opportunities for coaching, clarification, and earlier intervention. Reflection shifted attention away from convenience alone and toward usefulness and limitations, helping students become more deliberate in how they adopted or rejected GenAI as a resource.

Outcome 3: Modeling Mattered More Than Policy Language

Across iterations, modeled examples of effective and ineffective prompting appeared to shape student behavior more strongly than syllabus policy language alone. Simply including GenAI guidance in course policies did not seem to influence decision-making as effectively as demonstrating prompting practices in recorded lectures, critiquing AI-generated responses, and comparing GenAI suggestions directly against rubric expectations. Students seemed better able to evaluate GenAI when that evaluation was routinely modeled rather than described abstractly, reinforcing the importance of prompt awareness and critical thinking in classroom practice (Walter, 2024).

These observations across course iterations suggest that students are more likely to use GenAI strategically when instructors show them how to question it, compare it, and revise beyond it rather than merely warning them about misuse. Open disclosure practices and visible modeling fostered more direct conversation via emails and discussion forums about acceptable use than policy reminders alone. In this sense, instructional design became more influential than prohibition. The tension around GenAI did not disappear, but it shifted from suspicion and enforcement toward strategy and discernment.

Outcome 4: Evaluating GenAI Improved Broader Help-Seeking and Feedback Practices

One of the most important outcomes was that evaluating GenAI critically appeared to strengthen students’ evaluation of other forms of support as well. During the Assess Feedback phase, students compared GenAI feedback with rubric criteria, instructor comments, peer review, and writing center recommendations. This often produced productive skepticism as students began to recognize that no single source of feedback should be accepted uncritically, a practice consistent with human-AI collaboration models that emphasize the combined role of human judgment and artificial intelligence in developing AI literacy (Tzirides et al., 2024).

Several reflections suggested that once students practiced identifying mismatches between GenAI feedback and assignment expectations, they also became more deliberate in how they interpreted peer comments, instructor feedback, and rubric guidance. Students asked clearer follow-up questions, revisited assignment criteria more intentionally, and demonstrated greater awareness that feedback must be judged in relation to the task rather than accepted because it sounded confident or authoritative. In several cases, this evaluative work redirected students toward human-centered feedback from peers, instructors, or writing tutors, even when doing so required more time and effort. GenAI did not replace human support; in many cases, it sharpened students’ attention to it.

Outcome 5: Optionality Preserved Student Agency and Multiple Pathways for Success

Not all students chose to use GenAI, and the principle of optionality remained important for that reason. A small but consistent group of students opted out entirely, often because of ethical concerns, discomfort with GenAI tools, or a preference for relying on human feedback and more traditional academic supports. Since GenAI was positioned within all available supports rather than as a required pathway, these students were not disadvantaged by their choice. CLEAR continued to function through the same sequence of clarification, resource mapping, feedback comparison, and reflection, with tutors, librarians, peers, rubrics, and instructor feedback serving as primary supports.

This pattern reinforced the importance of treating GenAI as part of a broader resource ecology rather than as the defining feature of any online course. The instructional goal was never universal GenAI adoption, but stronger strategic decision-making. Optionality preserved learner agency while supporting the broader UDL principle that students benefit from multiple pathways for action and engagement. When students were asked to decide whether GenAI was appropriate rather than required to use it, the focus remained on resource literacy rather than compliance.

Across all five outcomes, the most significant shift was conceptual. CLEAR did not eliminate misuse or produce uniform behavior, but it made students’ resource decisions more visible, discussable, and analyzable. Instead of asking whether GenAI belonged in a writing course, the more productive question became whether students were learning to choose and evaluate available resources strategically all together. When that question drives instructional design, GenAI integration becomes less about control and more about skill building to be a successful online learner.

Conclusion

GenAI is often treated as the central problem in contemporary writing instruction, but in online first-year composition, the deeper instructional issue is resource literacy. Students have long had access to learning support, yet many still struggle to determine when support is needed, which resource is most appropriate, and how to evaluate whether feedback actually advances their goals. GenAI did not create this challenge; it made it more visible. Since GenAI tools are immediate, appear to be fluent, and easily mistaken for authoritative, they expose the broader need for students to learn how to compare resources critically rather than accept support passively. The most important pedagogical question, therefore, is not whether students should use GenAI, but whether they are developing the metacognitive habits and evaluative judgment necessary to use any resource strategically.

The cycle of CLEAR offers one practical model for addressing that challenge through instructional design rather than restriction alone. By embedding planning, resource selection, guided engagement, feedback comparison, and reflection into a repeatable cycle, CLEAR makes resource use visible and coachable across the writing process. Across course iterations, this approach appeared to support stronger reflection, more deliberate help-seeking, and greater willingness to question both AI-generated and human feedback. The value of the model lies less in the technology itself and more in the routine it creates for helping online community college students practice strategic decision-making over time.

For faculty and instructional designers, the implication is straightforward: GenAI integration should be tied to existing learning outcomes rather than treated as a separate policy issue. Assignment modules should include task-clarification prompts, resource-planning activities, guided comparison of feedback sources, and structured reflection opportunities as standard parts of course design. Faculty development should move beyond detection tools and policy debates to include practical models for transparency statements, prompting demonstrations, and feedback evaluation routines. Institutions do not need to adopt CLEAR as a fixed model, but they do need to design visible cycles of planning, engagement, assessment, and reflection so that resource use becomes teachable rather than assumed. As GenAI tools continue to evolve, the goal should not be to normalize GenAI uncritically or reject it categorically, but to help students build forms of self-regulation, resource literacy, and evaluative judgment that remain valuable beyond any single platform. The most sustainable response to GenAI is not better policing, but better teaching of discernment.