Advancing Translational Science Through AI-Enhanced Teacher Learning for Early Language and Composing

Abstract

Composing in early childhood classrooms offers a critical opportunity to strengthen children’s language skills, yet many teachers feel underprepared to provide this instruction. This study examines whether an AI-enhanced digital platform (L4C) can serve as a sustainable, community-based professional development model that bridges theory and practice. Twenty-nine teachers in the southeastern United States engaged with L4C, a professional learning model designed to integrate principles from the Science of Literacy, Learning, and Instruction into a cohesive platform that links teachers’ content and pedagogical knowledge-building with lesson planning and reflective practice. Data sources included surveys, pre- and post-lesson plans, and AI usage logs from the lesson planning tool. Findings showed that teachers initially reported significant barriers to composing instruction and sought professional learning responsive to their classroom needs. After using L4C, teachers demonstrated notable growth in their knowledge of language components and the quality of their composing lesson designs. Teachers evaluated the platform positively, particularly valuing the linked videos and scripted lesson tools for making theoretical concepts actionable. These findings suggest that AI-driven platforms like L4C can advance teacher learning in practical, individualized, and contextually relevant ways, offering a promising pathway for professional development in early literacy instruction.

1. Introduction

Early childhood classrooms provide a critical window of opportunity for shaping children’s language and literacy trajectories, thereby laying the foundation for later academic success (Dickinson & Tabors, 2001). A central goal during this period is fostering children’s facility with oral language that mirrors the register of school—often referred to as academic language—which resembles the language encountered in written texts (Farrow et al., 2024c; Rowe & Snow, 2020; Schleppegrell, 2012). Developing greater facility with academic language in the early years is important because it is strongly associated with later literacy and academic achievement (van Kleeck, 2015). Although many children enter school proficient in their home language, developing proficiency in academic language requires learning to adapt communication for audiences and purposes beyond familiar, everyday contexts. This shift entails increasing precision and specificity in language use as children express conceptual information, ideas, and explanations (Farrow et al., 2024c). This developmental shift depends on opportunities to use more complex linguistic forms that support the communicative purposes of schooling.

1.1. Importance of Early Writing to Build Facility with Language

One promising avenue for supporting children’s development of academic language is through early composing instruction (Farrow et al., 2024a). According to the Not-So-Simple View of Writing (Berninger & Winn, 2006), composing texts requires children to engage their linguistic resources while simultaneously maintaining awareness of both communicative purpose and content. This orchestration spans multiple levels of language. At the discourse and pragmatic level, children retrieve topic-relevant ideas, filter out irrelevant information, and sequence ideas to achieve coherence and elaboration. At the word and sentence level, they select from a constellation of linguistic forms to represent their intended meaning. Finally, to transcribe their ideas into print, children draw on knowledge of phonemes and their corresponding graphemes while coordinating language-dependent early literacy skills, such as directionality and letter formation, within the motor system.

Table 1. Language Components based on the NSSVW.

Domain	Skill/Target	Definition
Language-dependent early literacy skills	Letter Knowledge	Recognize and recall letters of the alphabet by name and form.
	Letter and Word Formation	Form individual letters legibly and fluently, and combine them to construct words, integrating fine-motor coordination
	Phonemic Awareness	Pronounce, isolate, segment, and manipulate sounds in spoken words.
	Sound–Letter Correspondences	Map spoken phonemes (sounds) to written graphemes (letters).
	Directionality	Understand and apply left-to-right, top-to-bottom sequencing in writing.
	Print Conventions	Apply writing mechanics (spacing, capitalization, punctuation).
Foundational oral language skills	Vocabulary	Breadth and depth of word knowledge to express ideas.
	Morphology	Use meaningful word parts (prefixes, suffixes, roots, inflections) to construct words accurately.
	Syntax	Form ideas using the hierarchical structure of language to coordinate and subordinate ideas.
Knowledge-dependent discourse/ pragmatic skills	Idea Initiation	Recall and generate knowledge or experiences related to a topic.
	Cohesion	Choose and connect relevant details while excluding irrelevant information.
	Coherence	Elaborate on a topic with details that build and link together logically.
	Genre Awareness	Identify and organize ideas according to genre-specific structures and features.

provides a detailed breakdown of early writing skills.

The complexity of the writing process directly engages children’s executive functions. In a recent longitudinal study, C. Puranik and Li (2022) demonstrated that early writing predicted gains in executive functioning, but not vice versa, suggesting that writing may serve as a pathway for strengthening cognitive abilities foundational to literacy and broader academic achievement. Similarly, Manfra et al. (2017), in a study of preschoolers from high-needs districts where the majority of children were African American, found that prekindergarten writing ability was the most robust predictor of third-grade academic achievement across multiple domains, even after controlling for other early literacy skills. The authors argued that writing’s integrative nature—requiring the coordination of motor, cognitive, and linguistic processes—may reinforce children’s developing literacy skills in powerful ways. Neuroimaging research supports this claim: children who practice writing letters, compared to those who only view them, show heightened activation in brain regions associated with deep processing and consolidation into long-term memory (James et al., 2017; Vinci-Booher et al., 2025). Intervention studies converge on this conclusion, showing that children who receive explicit early writing instruction outperform peers who receive equally strong instruction in reading or isolated early literacy skills, with advantages extending to phonemic awareness, letter–sound correspondences, and decoding (Aram & Biron, 2004; Berninger et al., 2002; Sénéchal et al., 2012). Taken together, this evidence suggests that writing’s heightened engagement of cognitive systems provides a powerful mechanism for accelerating literacy growth. Yet, literacy is fundamentally language-dependent. Letter knowledge, for instance, has little value apart from children’s ability to analyze and map it onto the phonological structure of spoken words. Writing, uniquely, is a language-output activity that requires children to make deliberate choices across multiple linguistic levels for the purpose of expression—a metalinguistic act that taxes working memory and compels conscious reflection on language (Fayol, 2017). From this perspective, writing is not only a pathway to accelerate foundational literacy skills but also a means of strengthening children’s language resources, particularly their emerging facility with academic language, which serves as the bedrock of later reading and writing achievement (Farrow et al., 2025b).

Prior research has consistently documented strong interrelations between oral and written language production, with evidence that children’s language abilities shape both the fluency and quality of their writing (Berninger et al., 2002; C. S. Puranik & Lonigan, 2012; Abbott et al., 2010). For example, Cabell et al. (2022) demonstrated that children’s growth in oral narrative skills predicted concurrent growth in both writing fluency and spelling, suggesting that advances in oral and written language unfold in tandem. Extending this line of work, Farrow et al. (2025a) showed that preschool children’s gains in translation abilities during composing tasks—indexed by measures of oral productivity, complexity, and relevance—predicted subsequent improvements in their oral narrative language. Together, these findings point to an integrated and potentially reciprocal relationship between oral and written language development, in which growth in one domain supports advancement in the other.

Crucially, experimental evidence shows that instruction targeting linguistic features of writing—such as word choice, sentence structure, and cohesive organization—improves children’s ability to generate and communicate ideas in writing during the early elementary years (Berninger et al., 2002; Harris et al., 2023; Klein et al., 2022; Traga Philippakos, 2019). Such findings suggest that writing instruction may provide a unique leverage point for accelerating language development, because it requires children to engage in conscious analysis and manipulation of language for communicative purposes. This metalinguistic demand positions writing as more than a tool for recording ideas; it is also a powerful instructional pathway for strengthening children’s facility with academic language. For children from vulnerable populations, who may have fewer opportunities to practice using language in decontextualized, academic ways, early writing instruction may therefore represent a critical window of opportunity to advance both language and literacy development.

1.2. The Problem: The Neglect of Writing in Early Childhood Classrooms

Despite the promise of early writing as a developmental pathway, research on classroom instruction paints a discouraging picture. Studies consistently find that the quality of writing instruction in early childhood classrooms is relatively low (Bingham et al., 2017; Coker et al., 2018; Kim et al., 2013). When writing opportunities are provided, they most often take the form of independent writing tasks with little to no targeted instructional support. Moreover, when teachers do engage children in writing, instruction is typically focused on isolated emergent literacy skills—such as naming letters or forming print—rather than on meaning-making or communicative purposes that activate children’s language across multiple levels (C. S. Puranik et al., 2014; Guo et al., 2023). For instance, Bingham et al. (2017) documented that preschool teachers overwhelmingly supported discrete skills, such as letter knowledge, while providing very limited guidance on composing skills that involve constructing and conveying messages to an audience.

This narrow focus reflects a broader issue: teachers often feel underprepared to teach writing and lack both the knowledge and support needed to integrate it meaningfully into their classrooms (Graham, 2019; Hindman & Wasik, 2008; McKenna & Schachter, 2025). Writing instruction requires understanding the constellation of subskills involved, how those skills interrelate, and how to intentionally target and scaffold them for young learners. Yet teachers’ knowledge of language, a central foundation for writing instruction, is often limited and is rarely a focal point of teacher preparation programs (Graham et al., 2022; McKenna & Schachter, 2025). Without this foundation, teachers understandably avoid or minimize writing instruction, leaving children with few opportunities to engage in the kinds of integrated language activities most critical for their development.

The consequences are particularly acute for children most vulnerable to later language and literacy difficulties. These students often require multiple, explicit, and systematic opportunities to use language in meaningful contexts to close persistent achievement gaps (Lonigan & Phillips, 2016). Unfortunately, the lack of high-quality writing instruction represents a missed opportunity to accelerate language growth and reduce these disparities. In response, professional development (PD) has been advanced as a critical mechanism for strengthening teachers’ knowledge of early writing and language skills—and for deepening their understanding of how to teach these skills effectively through evidence-based practices aligned with how children learn (Graham et al., 2001; McKeown et al., 2023; Moats, 1994, 2009). Yet, teachers’ ability to apply new knowledge often depends on whether professional learning feels accessible, relevant, and worthy of their limited time (Harris et al., 2023; McKeown et al., 2023; Owens, 2024). Moreover, sustained buy-in depends on whether PD recognizes and includes teachers as valued participants in the learning process (Reichenberg, 2022). L4C was intentionally designed to address these barriers, integrating these essential elements into a cohesive platform that makes professional learning both practical and transformative.

1.3. Response to the Problem: Designing the L4C Platform

To address persistent challenges in early childhood writing instruction, we developed Language for Composing (L4C), an AI-enhanced digital platform designed to advance the Language for Composing Method. Central to this method is the premise that writing development is fundamentally driven by language: children’s ability to compose text depends on their facility with oral language, language-dependent skills, and knowledge-dependent language skills operating in coordination (see Table 1). The approach is guided by four principles: (1) teachers must intentionally target specific components of language through explicit instruction; (2) instruction should maintain a level of linguistic intensity that pushes children toward more complex expression; (3) learning must be anchored in authentic communicative purposes, ensuring that language development is inseparable from meaning-making and content knowledge; and (4) transcription, language construction, and idea generation must be integrated, enabling children to engage with language across all levels of writing simultaneously.

The L4C platform operationalizes these principles through a framework grounded in multiple domains of scientific evidence. The Science of Literacy specifies the language skills children need to achieve literacy, with particular emphasis on writing. The Science of Learning explains how children acquire these skills through the interaction of working and long-term memory, emphasizing the need to design learning experiences that foster consolidation and transfer. The Science of Instruction identifies effective pedagogical practices and organizational scaffolds within and across lessons that optimize skill development. Finally, Translational Science addresses barriers to implementation by designing professional learning that is practical, feasible, and responsive to the needs of diverse classroom contexts. Together, these four domains form the foundation of our L4C approach design.

Within the digital platform, teacher learning is tightly coupled with classroom application, reflecting principles of practice-based PD (Farrow et al., 2024b; Grossman et al., 2009; Harris et al., 2023). L4C organizes writing instruction around three interrelated composing processes: transcription (handwriting and spelling), language construction (vocabulary, morphology, and syntax), and idea generation and organization (discourse and genre knowledge). For each composing process, teachers engage with brief, targeted videos that introduce the underlying language skills and their developmental foundations. Each composing process is paired with evidence-based instructional practices and explanations of how these practices support corresponding language skills (e.g., comparing and contrasting phonemes to strengthen transcription). To bridge theory and enactment, short classroom demonstration videos show how these practices operate dynamically within authentic classroom interactions. This combination of explicit knowledge-building and situated classroom modeling ensures that teachers not only learn the practices but also understand how to scaffold them effectively for diverse learners—in other words, how to teach in ways that align with how children learn.

A central innovation of L4C lies in its integration of professional learning and lesson design. Traditionally, teacher knowledge-building and instructional planning are treated as parallel but disconnected processes (Desimone, 2009). In contrast, L4C merges these processes so that teachers deepen their understanding of language-focused practices while simultaneously applying this knowledge through an AI-enhanced lesson planning tool. Teachers are prompted to create three coordinated lessons—one for each of the composing processes: transcription, language construction, and idea generation and organization. The tool then guides teachers in designing lessons that integrate these processes so that children can coordinate them for the purpose of written communication, thereby strengthening the foundational writing system. For example, a teacher might identify /p/ and the letter p as targets during transcription instruction. Using this input, the tool automatically configures opportunities for children to retrieve and apply the targeted transcription skill within corresponding language construction and idea generation activities. Similarly, when teachers select focal vocabulary or unit concepts, these are woven into sentence-building tasks and revisited in composing prompts, enabling children to apply their linguistic knowledge in meaningful contexts. Throughout the process, teachers are introduced to best practices across the composing components (e.g., providing opportunities for invented spelling) and can see how these strategies interact to support integrated, language-rich writing instruction.

Importantly, these lessons follow a consistent structure that routinizes instruction based on principles from the science of learning. For example, lessons across the three composing processes begin with a “Review & Connect” section designed to prompt retrieval practice—supporting children’s recall of previously learned concepts, processes, or skills and linking them to new content. Within each script, the rationale for every section is explicitly explained to teachers, clarifying its instructional purpose. In this way, teachers are continually reintroduced to practices that reflect how children learn and can see how evidence-based strategies fit coherently within the overall instructional framework. Crucially, the design connects directly to what teachers already do in their classrooms, enabling them to deepen their professional learning through their everyday instructional work. Rather than adding to teachers’ workload, the platform harnesses existing practice as the context for learning, making professional growth relevant, practical, and sustainable.

Another important component of the tool is its adaptability. Teachers can modify lessons to meet the diverse needs of their students—reducing cognitive load for learners who require additional support or increasing linguistic challenge for those ready to advance. These adaptations are made through interactive exchanges with the AI tool, allowing teachers to iteratively refine instructional plans. Once the teacher is satisfied with the adjustments, the system generates a revised lesson script tailored to individual learning profiles. In this way, lesson planning evolves into a structured opportunity for reflection, rehearsal, and pedagogical growth, transforming what is often a procedural task into a meaningful act of professional learning.

Finally, L4C functions as a dynamic, teacher-centered ecosystem rather than a static repository of resources. Teachers actively contribute to the refinement of the platform by uploading classroom videos, modeling enactment of practices, and sharing strategies with colleagues. This participatory design elevates teacher expertise and fosters a professional learning community in which teachers act as co-investigators of best practices. By embedding structured guidance, adaptable lesson design, collaborative knowledge-building, and AI-enhanced feedback, L4C not only strengthens teacher knowledge and instructional practice but also positions teachers as central drivers in the evolution of the tool. This responsiveness to teacher needs exemplifies translational science in action and ensures that L4C remains practical, rigorous, and sustainable for advancing children’s oral and written language development.

Given the vital role of technology in translational research designs aimed at accelerating children’s language growth, this study provides a preliminary investigation of L4C, an AI-enhanced PD tool designed to strengthen composing instruction. Specifically, we examine L4C’s potential to deepen teachers’ knowledge of the language processes that support early writing, to prompt reflection on their instructional practices, and to improve the quality of their lesson planning. In addition, we investigate the barriers teachers face in implementing composing instruction and how engagement with L4C may address these challenges. Our study is guided by four research questions:

RQ1. How do teachers perceive the support they currently receive for implementing composing instruction, and what barriers do they identify in their instructional contexts?

RQ2. To what extent do early childhood and elementary teachers understand the language components that underpin early writing, and how does this knowledge change after engaging with L4C?

RQ3. In what ways do teachers’ early writing lesson plans change after using L4C, particularly in terms of lesson focus, alignment with the Language for Composing method, and integration of evidence-based practices grounded in the science of learning?

RQ4. How do teachers interact with the tool, and which features are perceived as most effective in supporting changes to instructional planning?

2. Materials and Methods

2.1. Participants

Participants were 29 teachers from urban and suburban districts in a southeastern U.S. state. The sample was racially and ethnically diverse, with 15 teachers (52%) identifying as African American/Black, 8 (28%) as Hispanic, and 6 (21%) as White. All participants were enrolled in a master’s-level introductory course on literacy theory and instruction. Most (86%; 25 of 29) were teaching in early childhood classrooms (Pre-K through Grade 1). The remaining four (14%) were teaching in early elementary classrooms (Grades 2–3) but were included in the study because of their prior experience as early childhood teachers. All participants identified as female, with an average of 5–10 years of classroom teaching experience. Most held a bachelor’s degree (n = 26), or 90%, while two held a master’s degree and one held a doctoral degree. Approximately half (48%) reported extensive prior PD in structured literacy, whereas the remaining teachers (52%; n = 15) reported limited or no prior PD in this area.

2.2. Procedures

Before receiving any course instruction in writing, participating teachers completed a Qualtrics survey that collected demographic and background information and assessed their perceptions of both the supports available to them and the barriers they faced in delivering high-quality writing instruction. They also completed a brief questionnaire measuring their knowledge of language components. In addition, participants analyzed a kindergarten writing sample and developed a corresponding lesson plan to support the child’s writing development.

Following these pre-instruction measures, participants were asked to engage with the digital platform during two sessions of the course. Each class session was 2.5 h for a total of 5 h. During the first session, teachers were introduced to the AI-enhanced Language for Composing (L4C) web platform. A 10 min tutorial provided guidance on navigating the site’s resources and demonstrated the use of the AI-supported lesson planning tool. Resources included short instructional videos, spanning 5–8 min on the Not-So-Simple View of Writing (NSSVW), stages of writing development, the L4C approach, and exemplars of evidence-based practices for writing instruction.

Participants were then asked to engage independently with the L4C platform over a two-week period. The purpose of this phase was to deepen their understanding of the language-for-composing framework and its associated best practices for supporting early writing development. At the conclusion of this period, participants were provided with the original lesson plans they had created during the pre-instruction phase and were instructed to revise and/or create new plans based on their engagement with the platform. Revised plans were submitted the subsequent week. In addition, participants again completed a brief questionnaire assessing their knowledge of language components related to early writing. The study concluded with a final survey in which teachers reflected on the influence of the L4C platform on their instructional planning. The study spanned a total of five weeks. Importantly, during the three weeks in which teachers were independently engaging with the L4C platform, concurrent class sessions focused on culturally responsive literacy instruction and reading fluency and comprehension. Thus, no additional direct instruction related to early writing was provided during this period.

2.3. Measures

• Pre-survey Measure of Supports and Barriers to Writing Instruction.

Prior to course instruction, teachers completed a pre-survey designed to assess their perceptions of supports and barriers related to implementing high-quality writing instruction. The survey consisted of six items. Two items employed a 5-point Likert-type scale (1 = Not at all to 5 = A great deal). The first asked the extent to which teachers’ schools emphasized writing instruction, and the second asked the extent to which their undergraduate training had prepared them to teach writing.

Two subsequent items were multiple-choice questions, in which teachers could select multiple options and/or provide open-ended responses. These items asked teachers to indicate (a) what had helped them develop their writing instruction and (b) what had not been helpful. Example response options included “PD offered by my school, research I have done on my own, and disconnect between curriculum and how students respond to those lessons.” We present descriptives of all response options for Research Question 1 in the Section 3.

The final two survey items were open-ended questions that asked teachers to describe which components of PD they perceived as most helpful for improving their writing instruction and which they found less helpful. These qualitative responses were analyzed using an inductive open-coding process to capture the breadth of teachers’ perceptions regarding factors that facilitated or hindered their instructional growth. The analysis began with line-by-line coding of teacher responses to identify discrete and meaningful segments of text. Initial codes were descriptive and closely aligned with participants’ own language to preserve authenticity and avoid premature interpretation. Codes were refined through an iterative process of constant comparison (Glaser & Strauss, 1967), in which the research team examined similarities and differences across responses to generate abstract categories that reflected shared perceptions. To establish reliability, 10 teacher responses were randomly selected and independently coded by the first and second authors. Interrater reliability, calculated using Cohen’s κ, indicated substantial agreement between raters (κ = .87, p < .001; Landis & Koch, 1977). Once reliability was established, the remaining 19 responses were coded by the first author using the refined coding scheme. This second round of coding allowed us to apply the finalized categories consistently across the dataset and to capture the frequency with which each category was referenced by participants.

• Coding Pre- and Post-Teacher Knowledge of Language Components Supporting Early Writing.

To assess teachers’ knowledge of the language components that support early writing, we asked them—both before and after use of the L4C platform—to identify the language and language-dependent skills that children draw upon when composing texts. Teachers provided open-ended responses, enabling us to evaluate whether they could specify language skills relevant to sub-processes of writing (e.g., idea generation and organization, language construction, and transcription; Berninger & Winn, 2006). This coding approach captured not only the breadth of teacher knowledge but also the specificity and accuracy with which teachers identified the language components central to early writing development. The underlying rationale was that if teachers could not decompose the writing process into the discrete skills children need to execute these sub-processes, they would be unlikely to effectively target, scaffold, or monitor growth in those skills. In such cases, instruction would remain at a broad level rather than at the level of precision necessary to foster children’s language use in composing.

To analyze participant responses, we systematically coded each response for the presence or absence of the 14 language and language-dependent components (see Table 1). For example, a response such as “know their letters” was coded as letter knowledge. Repetitions of the same component within a single response were collapsed so that each component was counted only once per participant. This procedure allowed us to capture the specific language components teachers identified, rather than the frequency with which they repeated them. Using these presence/absence codes, we then calculated a proportion score for each participant, defined as the number of correctly identified components divided by the total number of possible codes (14). This provided a standardized measure of teachers’ knowledge of language components on a 0–1 scale.

In addition to coding correct responses, we also analyzed incorrect responses to capture teachers’ misconceptions about early writing. These responses were grouped into three emerging categories: (a) motor processes (e.g., “hold a pencil,” “fine-motor control”), (b) reading skills (e.g., “phonics,” “decoding”), and (c) broad processes (e.g., “revising,” “spelling”). This dual coding procedure allowed us to assess not only the accuracy of teachers’ knowledge of language components but also the types of misunderstandings that may shape their instructional practices. Incorrect response codes were quantified as raw frequencies to capture how often each category was mentioned across participants.

Following initial open coding by the first author, the second author independently applied the scheme to the full dataset. Discrepancies were resolved through discussion, and consensus codes were assigned. Interrater reliability between the first and second authors was substantial (Cohen’s κ = .89, p < .001). To quantify teachers’ knowledge, we calculated the proportion of correctly identified language-related components (i.e., language-based or language-dependent code-based) relative to the 14 possible correct components. Higher ratios indicated greater precision in teachers’ understanding of the linguistic skills that support children’s translation and transcription processes in writing.

• Coding Lesson Plans Before and After Engagement with AI Planning Tool.

Lesson plans were coded along three dimensions: (a) instructional focus of the lesson, (b) alignment with the Language for Composing approach, and (c) presence of evidence-based practices consistent with the Science of Learning. First, lesson plans created both with and without the AI tool were coded for instructional focus. Each lesson was classified into one of the three early writing processes: (a) transcription, (b) language construction, or (c) idea generation and organization. Within each process, component skills (e.g., sound-symbol correspondence for handwriting and spelling, vocabulary for language construction, coherence for idea generation) were further identified and coded. This level of coding allowed us to examine how teachers interacted with the AI Lesson Planning Tool by making visible the specific skills emphasized in their instructional planning. Descriptive statistics were then generated to compare teachers’ lesson foci across the tool-assisted and independent planning conditions. All component skill codes are reported in the Section 3.

Secondly, lesson plans were coded for their alignment with the Language for Composing method using an a priori scale. As outlined in the literature review, this method emphasizes supporting multiple dimensions of language synergistically as children compose texts. The coding scheme included four dimensions: First, we examined the targeted language focus, or the extent to which a specific language component was explicitly taught. Second, we assessed the level of language intensity, reflecting whether the language focus was calibrated to advance children’s current ability levels. Third, we coded for the interconnection between language and learning, which captured the extent to which language was used as a tool to help children recall previously learned knowledge and express that knowledge during composing. Finally, we examined the integration across levels of language, referring to the degree to which higher- and lower-level processes were coordinated in ways that supported their functional use together.

Each lesson plan was rated on a four-point scale (0 = not at all, 1 = a little, 2 = a moderate amount, 3 = a great deal). A score of 1 indicated an unintentional or disconnected representation of the principle; a 2 reflected emerging intentionality, though integration across principles remained limited; and a 3 indicated that the principle was fully integrated and intentionally represented within the lesson. Scores across the four dimensions were summed to yield a composite Language for Composing Method score.

Finally, lesson plans were coded for the presence of evidence-based practices consistent with the Science of Learning. Within our structured composing framework, we identified eight core practices: (a) reviewing previous material, which strengthens retrieval pathways from long-term memory and reduces cognitive load by activating prior knowledge; (b) connecting old to new, which supports integration of new content or skills with established schemas and facilitates transfer; (c) defining the skill or purpose, which promotes goal-oriented learning and clarifies instructional intent, thereby increasing the likelihood of effective working memory engagement (Willingham, 2009); (d) comparing and analyzing examples and non-examples, which helps students establish conceptual boundaries and refine schemas (Engelmann & Carnine, 2016); (e) modeling and explaining the skill or process, which provides a cognitive scaffold and reduces ambiguity during initial learning; (f) cueing application of the skill, which fosters transfer and metacognitive awareness by signaling when and how to apply knowledge; (g) providing specific corrective or confirmatory feedback, which helps students calibrate understanding and refine mental models, critical for long-term retention (Stockard et al., 2018); and (h) offering encouragement and praise, which sustains motivation and effort—factors that, while not directly cognitive, are essential for engagement and learning outcomes (Kim, 2020). Because encouragement and praise typically occur responsively during instruction rather than being pre-planned, this practice was excluded from the evaluation of lesson plans.

Each lesson plan was coded for the presence or absence of teacher actions reflecting one of the evidence-based practices. For instance, in the excerpt “Teacher will review what an adjective is with students,” the action was coded as reviewing previous material. The statement “Remind children that an adjective is a word that describes a noun” was coded as defining a skill. The prompt “Ask students to give examples of adjectives” was coded as reviewing previous material (retrieval). Finally, “Compares a simple sentence with no adjectives to one with adjectives” was coded as comparing and analyzing examples and non-examples. To quantify instructional quality, we calculated a standardized proportion score by dividing the number of practices coded as present by the total number of possible practices (7). A full list of the evidence-based practices is provided in Table 2 in Section 3.

Interrater Reliability for Lesson Plan Coding. To establish the reliability of the coding scheme, a graduate research assistant (GRA) with prior teaching experience was trained by the first author to apply the coding systems for the Language for Lesson Focus, Composing Method, and Evidence-based Practices. Training consisted of iterative review and discussion of sample lesson plans until coders achieved at least 90% agreement across five consecutive practice lessons. After training, the first author randomly selected 10% of the total lesson plans (N = 58) for independent double coding. Interrater reliability was assessed using Cohen’s kappa, which corrects for chance agreement. The resulting kappa coefficients were .865 for the Lesson Focus codes, .768 for the Language for Composing Method codes, and .722 for the Evidence-based Practices codes. According to benchmarks proposed by Landis and Koch (1977), these values represent substantial agreement, thereby providing evidence for the reliability of the coding procedures.

• Coding Teacher Interactions with and Perceptions of the AI-Enhanced Lesson Planning Tool.

Teachers’ use of the AI-enhanced digital platform was captured through system-generated logs and supplemented by their survey responses.

AI Log Data. Log files provided information on the frequency of iterations teachers engaged in while using the lesson planning tool. Iterations reflected the number of adaptations or refinements made to a lesson plan within a single session. Greater numbers of iterations were interpreted as evidence of teachers making more extensive adaptations to align instruction with student needs. For each teacher, the total number of iterations across sessions was summed to provide an index of tool engagement.

Teacher Survey. Following teachers’ use of the tool, they completed a brief survey that included both multiple-choice and an open-ended item. Closed-ended items asked teachers to indicate which feature of the digital platform they found most helpful, which aspect of the lesson planning tool was most useful, which aspect of the digital platform was most helpful in rehearsing or visualizing their instruction, and whether they anticipated using the tool again. Teachers were also invited to provide suggestions for improving the tool. The open-ended responses were qualitatively analyzed using the same constant comparison method described above for the pre-survey items (Glaser & Strauss, 1967). Descriptive statistics were used to summarize the multiple-choice responses, providing a complementary view of overall trends in teacher perceptions of the tool’s usefulness and areas for improvement.

3. Results

3.1. RQ1. Teacher Perceptions of Support and Barriers in Implementing Composing Instruction

The average rating of the emphasis schools placed on writing instruction was 2.3, suggesting that teachers generally perceived writing as receiving relatively little attention. In fact, 62% (n = 18) of teachers reported that their schools placed little to no emphasis on writing instruction. A smaller proportion, 21% (n = 6), indicated that their schools placed a moderate amount of emphasis, while only 17% (n = 5) felt their schools placed a lot to a great deal of emphasis on writing.

Interestingly, 21 teachers (72%) reported that their undergraduate training did not support their writing instruction at all, while six (21%) indicated it supported them a little, and only two (7%) felt it provided a moderate amount of support. Notably, no teachers reported that their undergraduate training prepared them a lot or a great deal. Overall, 93% of teachers perceived their undergraduate preparation as providing little to no support for teaching writing.

To further examine what teachers perceived as helpful for supporting their writing instruction, teachers were allowed to select multiple sources of support; therefore, percentages reflect the proportion of teachers endorsing each category. Eight teachers (28%) reported that their school’s PD supported them, whereas seven (24%) indicated that the PD they sought out independently was most useful. Nearly half of teachers (48%) noted that research they conducted on their own supported their instruction. In addition, 11 teachers (38%) identified their curriculum and related resources as supportive, while only four teachers (14%) pointed to their graduate studies. Overall, 62% of teachers reported that their writing instruction was primarily facilitated by research and PD they pursued independently.

In terms of factors that did not support their writing instruction, the majority of teachers pointed to challenges with curriculum and alignment. Eighteen teachers (62%) reported that a lack of quality curriculum and materials hindered their instruction, and 19 teachers (66%) indicated that a disconnect between curricular lessons and students’ responses limited their effectiveness. Additionally, five teachers (17%) noted that the inability to adapt lessons to student needs was unsupportive. Beyond curriculum, 12 teachers (41%) viewed their school-provided PD as unhelpful, 11 teachers (38%) felt that vague standards made writing instruction difficult, and five teachers (17%) identified administrators’ and specialists’ lack of knowledge as a barrier to effective instruction.

Finally, in examining teachers’ perceptions of PD experiences that best supported or hindered their instruction,

Table 2. Emergent Categories of Professional Development Perceived as Supportive or Unhelpful.

Effective PD for Instruction	Example Teacher Quotes	n (%)
Classroom Application	“directly tied to what I’m doing in the classroom”	27 (93%)
Explicit Examples of Practices	“should show exactly how it looks when teaching”	24 (82%)
Planning with Colleagues	“time to talk over and plan with other teachers”	20 (69%)
Ongoing Support	“should not be something that is one and done”	19 (66%)
Adaptations for Students	“need to show how this relates to specific kids who may need help”	12 (41%)
Visual Handout	“provide an outline to follow along”	5 (17%)
Non-Effective PD for Instruction	Example Teacher Quotes	n (%)
Lack of Relevance	“it has nothing to do with my kids and what they need”	26 (90%)
Generic Presentation	“doesn’t get into specifics” “too vague”	22 (76%)
Disconnect Theory–Practice	“too heavy on theory without showing how to teach it”	20 (69%)
PowerPoint Overload	“too much information at once” slides upon slides”	13 (29%)
No Implementation Support	“give you stuff then say go ahead and figure it out”	11 (38%)
Cookie-Cutter Resources	“resources that are too cookie-cutter and unhelpful”	3 (10%)

presents the emergent categories along with the percentage of teachers who endorsed each component. The most frequently cited supportive feature was PD that directly addressed the specific challenges and unique needs of their classrooms (93%). Nearly all teachers (n = 27) emphasized that PD should be “directly tied to what I am doing in the classroom.” The next most prominent feature, reported by 82% of teachers (n = 24), was being explicitly shown how to implement a practice or strategy. In contrast, the majority of teachers identified irrelevant PD as most unhelpful (n = 26, 90%), followed by PD that was overly generic, focusing on “surface-level” content that was too vague to be translated into practice (n = 22, 76%).

3.2. RQ2. Change in Teacher Knowledge of Early Writing Language Components

Prior to using the L4C Platform, teachers demonstrated limited knowledge of the language and language-dependent components associated with early writing. On average, they correctly identified only 16% of the 14 components (M = 0.16, SD = 0.20; approximately 2.23 correct responses), while listing nearly five incorrect components (M = 4.70, SD = 2.40), most of which reflected overly broad processes (M = 3.87, SD = 2.10).

After using the platform, teachers demonstrated substantial gains in knowledge. Accuracy increased from 16% at pretest to 45% at posttest (M = 0.45, SD = 0.22), t(28) = 11.15, p < .001, Cohen’s d = 2.02, a very large effect. Teachers’ average number of correct responses rose to 6.40 (SD = 2.30), while the frequency of incorrect responses declined significantly, t(28) = 5.78, p < .001, Cohen’s d = 2.05, with errors still most often reflecting overly broad processes. McNemar’s test further showed that the proportion of teachers correctly identifying specific components—including phonemic awareness, letter–sound correspondences, letter formation, vocabulary, syntax, and idea initiation—significantly increased (all ps < .05; see

Table 3. Proportions of Teachers’ Knowledge of Early Writing Language Components and Shifts Following Use of the L4C Platform.

Language Component	Pre-Use n (%)	Post-Use n (%)	Test of Change
Letter knowledge	7 (24%)	10 (34%)	χ2(1) = 1.9
Letter formation	3 (10%)	22 (76%)	χ2(1) = 15.4 **
Phonemic awareness	3 (10%)	18 (62%)	χ2(1) = 12.5 **
Sound–letter correspondence	4 (14%)	12 (41%)	χ2(1) = 4.1 +
Directionality	2 (7%)	7 (24%)	χ2(1) = 3.1
Print Conventions	12 (41%)	17 (59%)	χ2(1) = 1.8
Vocabulary	13 (44%)	26 (89%)	χ2(1) = 10.1 **
Morphology	2 (7%)	5 (17%)	χ2(1) = 0.9
Syntax	10 (35%)	18 (62%)	χ2(1) = 5.8 *
Idea initiation	6 (21%)	22 (76%)	χ2(1) = 16.1 **
Cohesion	0 (0%)	3 (10%)	---
Coherence	2 (7%)	5 (17%)	χ2(1) = 0.9
Genre features	3 (10%)	7 (24%)	χ2(1) = 2.3
Genre structure	7 (24%)	17 (59%)	χ2(1) = 5.8 *
Incorrect Component Categories	Pre-Mean (SD)	Post-Mean (SD)	Test of Change
Broad processes	4.06 (2.1)	1.97 (1.1)	t(28) = −5.8 **
Reading skills	0.33 (0.31)	0.22 (0.34)	t(28) = −0.7
Motor processes	0.45 (0.39)	0.23 (0.43)	t(28) = −1.1

Note. * p < .05, ** p < .01, + p < .09.

).

3.3. RQ3. Impact of L4C on Teachers’ Lesson Focus, Principle Alignment, and Instructional Plan Quality

Prior to being introduced to the L4C platform, teachers’ lesson planning largely emphasized surface-level print conventions. Nearly half of the teachers (n = 13, 45%) focused on aspects such as capitalization, correct spacing, and punctuation at the end of sentences. A smaller proportion focused on handwriting (n = 5, 17%), whereas no teachers directed attention to spelling. Four teachers (14%) emphasized sentence structure, which was generally limited to ensuring that students produced a complete sentence with a subject and verb. Vocabulary instruction was absent across all lessons (0%). Six teachers (21%) highlighted idea initiation, while only one teacher (3%) targeted broader aspects of structure, such as sequencing within a narrative.

Following use of the L4C platform, teachers’ instructional focus broadened substantially beyond surface-level transcription skills. No teachers (0%) emphasized print conventions, and only two teachers (7%) focused on handwriting. By contrast, five teachers (17%) addressed spelling, which often incorporated handwriting as a supporting component. Sentence-level instruction increased, with seven teachers (24%) emphasizing sentence construction. Vocabulary instruction also emerged, with two teachers (7%) incorporating word-level work that intersected with sentence development. Notably, attention to genre structure rose to eight teachers (28%), while five teachers (17%) focused on idea initiation.

As shown in

Table 4. Descriptives and Tests of Change in Teachers’ Alignment to L4C Principles and Instructional Quality.

Component of Lesson Plan	Pre-L4C		Post-L4C
	Mean (SD)	n (%)	Mean (SD)	n (%)	Test of Change
L4C Principles	1.2 (1.5)	--	5.2 (2.1)	--	t(28) = 8.2 **
Targeted language focus	0.97 (0.76)	--	1.81 (0.61)	--	t(28) = 5.7 **
Level of language intensity	0.10 (0.31)	--	1.2 (0.87)	--	t(28) = 6.4 **
Interconnection	0.24 (0.45)	--	0.90 (0.91)	--	t(28) = 3.1 *
Integration	0.31 (0.36)	--	1.30 (0.70)	--	t(28) = 6.2 **
Quality Practices	0.33 (0.15)	--	0.53 (0.17)	--	t(28) = 4.6 **
Review	--	15 (52%)	--	10 (34%)	χ2(1) = −1.2
Connect	--	4 (14%)	--	15 (52%)	χ2(1) = 2.6 **
Define	--	3 (10%)	--	5 (17%)	χ2(1) = 0.71
Examples	--	5 (17%)	--	10 (34%)	χ2(1) = 1.5
Model & Explain	--	26 (89%)	--	29 (100%)	χ2(1) = 1.3
Application	--	16 (55%)	--	26 (89%)	χ2(1) = 3.0 **
Feedback	--	3 (10%)	--	11 (37%)	χ2(1) = 2.3 *

Note. * p < .05, ** p < .01.

, teachers’ lesson planning demonstrated significant improvement following use of the L4C tool. Prior to implementation, lesson plans reflected an average of only 1.2 principles (SD = 1.5), indicating that most teachers incorporated a single evidence-based principle across their plans. Targeted language focus was the most frequently included principle at pretest (M = 0.97, SD = 0.76), though at a level suggesting limited, nonsystematic attention to language components. After engaging with L4C, teachers’ average number of principles increased to 5.2 (SD 2.1), representing a statistically significant gain, t(28) = 8.2, p < .001, d = 2.10. Paired-sample t-tests further indicated that teachers’ scores improved significantly across all four L4C principles (ps < .004).

Results for lesson plan quality followed a similar pattern. At baseline, teachers included 33% of possible quality practices in their plans; this proportion rose to 53% post-use, a statistically significant gain, t(28) = 4.6, p < .001, d = 0.71. Notably, Modeling & Explaining was the most consistently represented practice, with 89% of teachers including it at pretest and 100% at posttest; however, this increase was nonsignificant according to McNemar’s test, χ²(1, N = 29) = 1.3, p > .342. By contrast, practices requiring deliberate articulation of skill application, such as Define Skill and Use of Examples/Non-Examples, did not show significant pre–post gains (all ps > .05). Nonetheless, teachers did reflect increases in engaging children to apply language skills during activities, which in turn led to increases in their preparation to provide feedback on children’s application.

3.4. RQ4. Teacher Interaction & Perceptions of L4C Efficacy

Regarding the usefulness of the digital platform, most teachers (86%) reported that they would continue to use it. However, teachers varied in the specific features they perceived as most supportive of their instructional practice. A small proportion (n = 4, 14%) identified the explanation and organization of the platform into the three processes of early writing as most beneficial. More commonly, teachers pointed to concrete instructional resources, with 10 teachers (35%) highlighting the videos of best practice and 6 teachers (21%) emphasizing the organization and explanation of lesson components within the scripted lessons. Nine teachers (31%) indicated that the scripted lesson plans themselves were most useful. Overall, more than half of the teachers (52%) identified the scripted lesson feature as the most supportive component of the platform.

When asked which aspects of the AI-enhanced lesson planning tool most supported their ability to visualize instructional practice, teachers most frequently cited the scripted activities (n = 9, 31%). Nearly as many emphasized the value of modifications tailored to children’s ability levels (n = 8, 28%), which they reported helped them envision implementation more fully. Others highlighted the structuring of lessons into key component parts, which clarified how each segment supported and built children’s learning (n = 7, 24%). A smaller proportion indicated that examples of activities (n = 5, 17%) were most helpful.

Lastly, teachers provided input on potential improvements to the AI digital platform. More than half of the participants (n = 16, 55%) reported that the ability to generate printable materials would be particularly useful for supporting early writing instruction. Nearly half (n = 14, 48%) expressed interest in integrated assessments that could provide information about children’s developmental progress. In addition, one-third of teachers (n = 10, 34%) recommended including descriptions and summaries of the activities within each lesson to further enhance usability (see

Table 5. Teacher Suggestions for Improvement of the L4C Digital Platform.

Suggestions for Improvement	Teacher Quotes	n (%)
Ready to use materials	“download and print out lesson materials”	16 (55%)
Assessment	“link to an assessment to measure language”	14 (48%)
Activity description	“provide a full summary of the activity not just example”	10 (34%)
Link videos to script	“have videos linked right to practice in the script”	8 (28%)
Increase the age band	“should include older grades as well”	6(21%)
Configuration for ESOL & SWD	“should have a mode so that everything is for ESOL”	5 (17%)
Connect to district curriculum	“hook the system to our digital curriculum”	2 (7%)
Strategies for behavior management	“include how to deal with behavior problems”	2 (7%)

).

Finally, we examined teachers’ interactions with the AI lesson planning tool to assess the extent to which they adapted their lessons and, by extension, engaged in learning through the process. On average, teachers revised their lesson plans 5.9 times (SD = 2.4, Range = 2–12), indicating that they typically engaged in multiple cycles of adaptation with the tool.

Follow-up Analyses: Patterns between Variables. As a follow-up analysis, we explored whether patterns emerged between teachers’ lesson foci, language knowledge, quality of practices, and their iterations with the AI tool. To do so, we collapsed the seven specific lesson focus codes into three dummy-coded categories corresponding to the core processes of early writing (handwriting/spelling, vocabulary/sentence structure, and idea generation/organization). Pearson correlations revealed that, following L4C use, teachers with higher language component scores also demonstrated stronger alignment with the L4C instructional approach (r = .684, p < .001) and engaged in more iterations with the AI tool (r = .601, p < .001). In addition, quality of practices was significantly associated with teachers’ instructional focus. Teachers who emphasized vocabulary and sentence structure demonstrated lower quality scores (r = –.356, p = .034) and lower alignment with the L4C approach overall (r = .411, p = .031), particularly with respect to the intensity of instruction (r = –.404, p = .027). By contrast, teachers who focused on idea generation and organization exhibited higher quality scores (r = .545, p = .002) and stronger alignment with the L4C approach, especially through greater integration of reading–writing connections (r = .641, p < .001). Teachers’ language knowledge was not significantly related to practice (r = .261, p = .110); however, practice was positively associated with the L4C approach (r = .401, p = .027).

4. Discussion

The purpose of this pilot study was to evaluate the efficacy of the AI-enhanced digital platform, L4C, in supporting teachers’ planning of composing lessons that foster both oral and written language development in early childhood classrooms. A key innovation of L4C is its integration of teacher perspectives, allowing the platform to be continually adapted to the realities and challenges of classroom practice. As a potential PD ecosystem, it is essential to determine what forms of support most effectively enhance teachers’ instructional quality and, in turn, positively influence student outcomes. Accordingly, this study examined teachers’ perceptions of their current support for implementing composing practices, the barriers they encounter in doing so, and the professional learning opportunities they believe would strengthen their instruction. In addition, we assessed whether engagement with L4C enhanced teachers’ understanding of the language components underpinning early writing, improved the quality of their lesson planning, and provided a practical resource for ongoing instructional support.

4.1. Teacher Support, Barriers, and PD Needs

The findings from Research Question 1 highlight a substantial gap in the support available to teachers for early writing instruction. Most teachers reported that neither school- nor district-level initiatives nor teacher training programs adequately prepared them to teach writing. As a result, many indicated that they had resorted to conducting their own research and paying for PD opportunities out-of-pocket in order to strengthen their practice, aligning with prior findings (McKenna & Schachter, 2025; Graham et al., 2022). Teachers also expressed dissatisfaction with the curricula and instructional resources currently available to them, describing these materials as insufficient for guiding high-quality early writing instruction. Together, these reports underscore an urgent need to provide more robust, systematic support for teachers’ writing instruction, particularly in the early years of schooling.

The limited attention to writing in professional preparation stands in sharp contrast to the national momentum surrounding the “science of reading” in early literacy instruction. Although advances in reading have garnered substantial investment and widespread adoption in recent years, writing has remained peripheral in both research and practice, despite its well-documented role in reinforcing reading development (Graham, 2019; Kim, 2020) and its unique potential to accelerate cognitive-linguistic processes that underlie later academic success (Berninger et al., 2017; Manfra et al., 2017). Neglecting early writing instruction, therefore, represents a critical missed opportunity to build foundational skills, particularly for children who are most vulnerable to later literacy difficulties. These findings are consistent with prior work documenting teachers’ lack of preparation for writing instruction, their limited confidence in teaching it, and the corresponding trend toward minimal or low-quality writing opportunities in early childhood classrooms (Coker et al., 2018; Kim et al., 2013; C. S. Puranik & Lonigan, 2012).

At the same time, teachers’ responses also revealed a strong desire to improve their practice. Many reported actively seeking out early writing resources and PD opportunities on their own—an observation that aligns with earlier research pointing to teachers’ motivation to strengthen their knowledge when high-quality support is unavailable (Cutler & Graham, 2008). The most prominent source of dissatisfaction among teachers, however, was the lack of specificity in existing curricula and PD. This finding aligns with Wijekumar et al.’s (2019) content analysis of elementary writing curricula and textbooks, which revealed that these resources often fail to provide explicit guidance to support teachers’ knowledge of writing instruction. In many cases, the materials reflected misconceptions and practices misaligned with research-based best practices. Consistent with these findings, teachers in our study expressed a strong desire for concrete, explicit guidance on what to teach and how to teach it, rather than abstract strategies or generalized advice.

Across their accounts, a central theme emerged: teachers value translational practice. Specifically, they wanted to see how theoretical principles could be broken down into classroom-relevant practices, demonstrated in concrete ways, and situated within the realities of their own instructional contexts. These perspectives resonate with practice-based models of PD, which seek to bridge theory and classroom practice by centering instructional practices that are both observable and functional (Grossman et al., 2009; Harris et al., 2023). From this perspective, practices are not isolated techniques but purposeful actions that can be monitored for their impact on student learning. Teachers’ dissatisfaction with vague or decontextualized guidance suggests that their limited support may not simply reflect a lack of resources, but rather the scarcity of translational research that disaggregates the complexity of writing into meaningful components, concrete practices, and functional links to student outcomes. This gap has direct implications for the design of teacher learning opportunities. Specifically, it underscores the need for PD and instructional models that explicitly connect knowledge of writing processes to actionable classroom practice—a point we address in the next section.

4.2. Knowledge of the Language Components: Connections to Teacher Frustrations with Writing

We found that teachers’ knowledge of the component language and language-dependent skills involved in early writing was relatively low. This aligns with prior work showing that teachers’ knowledge of writing tends to lag behind their knowledge of other early domains, such as storybook reading vocabulary and emergent literacy skills (Hindman & Wasik, 2008; Bingham et al., 2022; Moats, 1994). Our results add an important layer to this literature by suggesting why writing knowledge may be comparatively weaker: teachers often lack explicit understanding of the language-specific components that underlie writing processes. Without this knowledge, teachers may struggle to target skills with precision, scaffold children’s application of those skills, or feel confident in their ability to teach writing effectively. Teachers’ responses illustrated this challenge. Many were able to name broad processes, such as revising, but struggled to identify the language resources children must draw upon to engage in those processes. This suggests that teachers understand writing at a procedural level but lack awareness of the specific linguistic underpinnings—such as vocabulary, syntax, or cohesion—that children coordinate when composing. The absence of this granular knowledge may prevent teachers from recognizing how component skills interact during writing, thereby reinforcing low confidence in instruction.

Although teachers’ knowledge increased significantly and meaningfully over the course of their engagement with L4C, their average identification of components still fell below half of the possible skills. This underscores both the inherent complexity of writing and the steep learning curve teachers face in mastering its language foundations.

Notably, teachers demonstrated relatively stronger growth in identifying transcription-related language skills (e.g., phonemic awareness, letter–sound correspondence). This likely reflects the overlap of these skills with reading instruction, where teachers already have more established knowledge, making them more familiar and accessible. At the same time, this finding highlights the importance of not only recognizing individual skills but also understanding how these skills interact within broader writing processes. In contrast, teachers showed less certainty about higher-order discourse processes, which are less explicitly emphasized in early curricula and thus remain more challenging to identify and target.

Taken together, these results suggest that PD must go beyond general encouragement to “teach writing” and instead provide explicit breakdowns of the language components that constitute writing processes. Doing so helps teachers move from a broad procedural understanding toward a more analytic and functional perspective, where specific language skills can be targeted, scaffolded, and monitored for student growth. The gains observed here, even over a relatively short period, are promising. They indicate that teachers can develop this knowledge when supported with tools that connect component skills to concrete classroom examples. This suggests that L4C represents a viable pathway for deepening teachers’ knowledge of the language foundations of writing, thereby equipping them to deliver higher-quality instruction.

4.3. Quality of Instructional Planning: Movement Toward Integration for Communication

Some of these same patterns were mirrored in teachers’ growth in both their use of the L4C approach and their incorporation of best practices in lesson planning. After engaging with L4C, teachers demonstrated notable shifts in the focus of their planned lessons. Specifically, they moved away from emphasizing isolated, code-related skills toward a more integrated set of language-based practices that directly support meaning-making in writing. For example, lesson plans showed less emphasis on print conventions and handwriting in isolation, and more attention to spelling—an area that inherently draws on phonemic awareness, letter–sound correspondences, and application of letter formation and directionality. Similarly, teachers began to address word choice (i.e., vocabulary) in the context of sentence construction, and some even targeted higher-level discourse skills, suggesting a growing effort to situate writing within more authentic communicative tasks. These shifts suggest that teachers were increasingly able to reflect L4C’s principles in their planning by focusing on targeted skills that, when integrated, elevate the linguistic intensity of instruction. Growth across all four principles also points to their interconnectedness, such that advancement in one principle appeared to facilitate progress in others.

Teachers’ reflections of best practices in their lesson plans followed a parallel but more nuanced pattern. Before engaging with L4C, nearly all teachers emphasized modeling, often in the form of the familiar “I do, we do, you do” approach widely promoted through district curricula and PD aligned with Science of Reading mandates. Although modeling remained a central practice, it was frequently represented as a one-time demonstration rather than as a recursive, scaffolded process that gradually fades as children assume greater responsibility. This reflects a broader limitation in teachers’ knowledge: the absence of a more comprehensive understanding of how language skills develop within and across components. Such understanding is necessary for greater precision in identifying (a) advanced language features (e.g., phrasal and clausal structures and their functions) and (b) the ways these features are orchestrated to support writing processes. Consistent with this limitation, teachers demonstrated less growth in explicitly planning to define concepts or to provide examples and non-examples, further underscoring these challenges. This may help explain why, in follow-up analyses, teacher quality of planning was not directly associated with measured knowledge of language components. It is possible that our assessment of teachers’ language knowledge was too cursory, failing to capture nuanced understanding of linguistic structures that might be most relevant for supporting high-quality instruction. Previous research has similarly noted limitations in measures of language and literacy knowledge, emphasizing that without precise instruments, it is difficult to identify which aspects of teacher knowledge most strongly influence instructional effectiveness. These findings align with Moats’s (2009) argument that literacy teachers require more rigorous, in-depth training in language, as all literacy activities fundamentally depend on strong language proficiency.

Building on these findings, follow-up analyses confirmed that instructional quality in teacher planning was closely tied to teachers’ instructional foci, suggesting that knowledge may exert its influence on practice indirectly through planning. Teachers who concentrated on vocabulary and sentence structure often demonstrated lower-quality instruction, likely reflecting limited understanding of how to sequence and connect linguistic components—for instance, linking denotation and connotation or extending syntactic structures beyond simple subject–verb propositions. Without this depth of knowledge, teachers may struggle to engage children in ways that foster advanced linguistic competence or to illustrate the functional purposes of these skills. This is particularly consequential given that vocabulary and syntax form the foundation for expressive facility and knowledge building, underscoring the need for stronger professional learning in these domains. In contrast, teachers who emphasized discourse-level skills demonstrated higher instructional quality, often through attention to idea initiation and genre organization. This pattern suggests that teachers are more accustomed to supporting brainstorming and outlining—likely reinforced by curricular materials—while vocabulary and syntax remain less explicitly addressed within writing instruction and knowledge construction. As a result, teachers with a stronger grasp of language knowledge may be better equipped to plan instruction, as they can more effectively break down concepts, provide clear definitions, and illustrate skills through examples and non-examples—capacities that are often diminished when such knowledge is lacking.

Nonetheless, we found significant increases in teachers’ inclusion of application and feedback opportunities, which indicates growing recognition of children’s active engagement with skills and of their own instructional role in scaffolding that application. This suggests that teachers’ plans included not only guiding skill use but also supporting children in applying those skills to express their ideas. Prior research highlights the importance of engagement in writing as a critical mechanism for coordinating higher-order cognitive and linguistic processes that contribute to children’s language development (Farrow et al., 2024a; James et al., 2017; Vinci-Booher et al., 2025). With continued exposure to L4C, teachers appeared to move beyond formulaic modeling toward more systematic and dynamic instruction that better supports the iterative nature of writing development. The platform provided a structured space for rehearsal, reflection, and access to resources, thereby facilitating the integration of professional learning with instructional practice. Follow-up analyses further indicated that L4C reinforced teachers’ application of its guiding principles, which in turn were associated with their language knowledge and the quality of their lesson plans. This suggests that the platform not only supported teachers’ understanding of its core tenets but also demonstrated the broader value of principle-based frameworks in helping educators connect theory with practice (Kavanagh et al., 2024). Overall, engagement with L4C enhanced teachers’ incorporation of quality practices aligned with how children learn, highlighting the importance of intentional design features in professional learning tools.

4.4. Supporting Teacher Growth in Translating Knowledge to Practice

In regard to teachers’ use of the L4C tool, we found that teachers generally engaged with the platform to adapt lessons to the particular needs of their students, suggesting that the tool can support teacher reflection and, by extension, their learning, particularly with respect to lesson organization and responsiveness to unique classroom contexts. Follow-up analyses revealed that teachers’ knowledge of language components after using L4C was related to the number of iterations they engaged in with the AI tool. This finding suggests that teachers who demonstrated broader understanding of language components across levels (e.g., subword, word, sentence, discourse) were more likely to engage repeatedly with the tool, whereas those with more limited knowledge engaged less frequently. Such a pattern implies that use of AI technology in instructional contexts may be partially contingent on teachers’ content knowledge, as that knowledge supports deeper reflection on how to present information to children at different developmental levels. Consequently, developers of AI-based tools for teachers must account for variability in teacher knowledge when designing platforms that aim to be both accessible and generative.

Finally, we investigated whether the design features of L4C supported teachers in the way we had hypothesized. Nearly all teachers reported that they would engage with the platform again, suggesting that L4C was viewed as instructionally valuable rather than burdensome or irrelevant to classroom needs. Teachers frequently highlighted the practice-based videos and scripted lessons as especially useful, as these concrete exemplars made abstract principles visible in practice and helped them connect theory to how instruction supports children’s learning. Prior work has underscored the importance of such representations for bridging the persistent gap between PD and classroom enactment (Desimone, 2009; Grossman et al., 2009; Harris et al., 2023). Teachers also emphasized the value of modifying lessons within the platform, which they described as both a reflective exercise and an opportunity to anticipate how instruction could be adapted to the diverse needs of their students. This aligns with research showing that reflection and adaptability are critical mechanisms through which professional learning influences instructional quality (Biesta et al., 2015; Loughran, 2002; van Es & Sherin, 2008). Collectively, these findings indicate that L4C design features supported teacher learning by offering concrete, adaptable, and practice-oriented resources—characteristics consistently identified in the literature as essential for meaningful and sustainable professional growth (Borko et al., 2008; Gaudin & Chaliès, 2015; Darling-Hammond et al., 2017).

Beyond exploring the feasibility and usability of the platform, we also sought teacher feedback on features that could enhance its efficacy in supporting instruction. Many teachers expressed a desire for the platform to include integrated materials that could be easily printed for classroom use, as well as an assessment function that would provide insights into children’s developmental levels to better guide instruction. These thoughtful responses highlight the potential for the platform to serve as an evolving ecosystem. Within this context, teacher needs and classroom contexts are directly connected to research aims, ultimately fostering stronger researcher–teacher–community partnerships.

In response, we are currently incorporating these adaptations into the L4C platform. Teachers will be able to download and print personalized materials for their lessons, such as subject–verb cards aligned with their unit topic or lists of words that begin or end with a target sound (e.g., animals in the forest that start or end with /m/). We are also developing an updated iteration of the platform that captures children’s oral and written language production in real time, providing teachers with actionable data to inform instruction. These enhancements further strengthen the connection between theory, planning, and practice, ensuring that teachers receive comprehensive PD that meaningfully supports their everyday instructional experiences. In addition, user-friendly resources are being developed to strengthen teachers’ understanding of sentence structure and vocabulary while extending support to higher-level discourse skills—such as coherence, cohesion, and genre-specific features—that have traditionally been less accessible in early writing instruction. By framing discourse-level skills as integrally connected to vocabulary and syntax, the platform makes these dimensions of writing more transparent and teachable for early childhood educators. These knowledge-building resources will be embedded as “More to Explore” sections within the transcription, language construction, and idea generation/organization modules. Based on each teacher’s interactions with the platform, specific resources will be recommended to address individual learning needs, providing targeted, multimedia supports that teachers can access as needed to deepen their knowledge and instructional practice.

4.5. Limitations of the Study

This study is not without limitations. First, we were unable to capture impacts on teacher practice or student outcomes; therefore, we cannot determine whether L4C directly influenced instructional implementation or children’s learning. While the observed gains in teachers’ reflective planning and language knowledge do not provide evidence of changes in classroom practice, they nonetheless indicate meaningful growth in the desired direction. Future research will therefore focus on examining whether engagement with L4C leads to measurable improvements in teacher practice and, subsequently, in children’s language and early writing outcomes. To this end, the next iteration of the platform will include features that capture child oral and written language data, enabling a more direct examination of how teachers’ engagement with L4C may support children’s language development. Additionally, the modest sample size limits the use of causal language, and we interpret these findings as indicative of the platform’s potential and as a pathway toward improving its efficacy. Importantly, this work revealed that teachers’ limited knowledge of early writing’s language components—and how these components interrelate—necessitated adapting the AI platform to include more embedded knowledge-building opportunities and personalized learning pathways. While these enhancements were not yet implemented in the present study, the findings nonetheless provide a critical foundation and direction for advancing L4C to better support teacher knowledge of the language components of early writing.

5. Conclusions

Early writing represents a critical window for strengthening children’s language abilities in ways that set the foundation for later academic success and broader well-being. Yet, the realities of classroom pressures often leave teachers with limited capacity to devote sustained attention to writing instruction, despite its importance. The present study demonstrates the promise of leveraging advances in AI technology to bridge this gap by translating research into practice through scalable, responsive professional learning. Our findings suggest that teachers engaged with the L4C platform not only to enhance their instructional planning, but also to integrate guiding principles of effective instruction—particularly those that foreground children’s use of language in meaningful communicative contexts. This pattern indicates that teachers recognized the practical significance of early writing and were motivated to continue learning in ways directly tied to their classroom practice. Importantly, the individualized and adaptive nature of AI-enhanced PD may provide unique advantages over traditional, one-size-fits-all models by making professional learning more relevant, efficient, and responsive to teachers’ instructional needs. Finally, the iterative design of L4C—continually refined in response to teacher feedback—positions the platform as more than a delivery mechanism for PD. It fosters a sense of professional community in which teachers both contribute to and benefit from collective knowledge-building. In doing so, L4C models a new generation of PD that integrates theory, planning, and practice in ways that not only save teachers valuable time but also enhance their instructional confidence and capacity to support children’s writing development.