1. Introduction and problem

In the 3-to-6 years band — kindergarten, preschool, CENDI, infant school — a package of three artefacts now claims to count as everyday lesson planning. The first is a model-generated lesson plan: a text with objectives, materials, timings and a “closing” that a system produces from a prompt. The second is a “ready-made” sequence: a downloadable package or catalogue lesson copied into the planning notebook. The third is a prompt that produces the day’s assembly: an instruction that asks for a song, a story and a circle, and delivers a script. All three are visible and cheap in coordination time. They allow a centre to exhibit that it “already plans with artificial intelligence.” The enunciative leap is enormous: one moves from generating a document to affirming that there is pedagogical planning. That leap is authorized neither by the evidence of how a plan is produced with a model nor by what early childhood frameworks call planning.

The thesis of this article is restrictive. A model-generated lesson plan, a “ready-made” sequence or a prompt that produces the day’s assembly does not constitute pedagogical planning. Planning in early childhood education is observation of the group, situated decision and in-the-act adjustment; not a textual artefact generated a priori. NAEYC locates observation as a process for planning and implementing, and states that educators use that information both in curriculum planning and in moment-to-moment interactions (NAEYC, 2020). Mexico’s National Curriculum for Initial Education describes planning that is integral, open, flexible and narrative: observe, record and propose environments (Secretaría de Educación Pública [SEP], 2024). Aistear, updated in 2024, promotes a cycle of noticing, nurturing, responding and reflecting, and distinguishes long-, medium- and short-term planning together with spontaneous planning (Government of Ireland, 2024). None of those frameworks authorizes treating a text generated before seeing the group as the act of planning.

The problem is sharpened by a professional reason that product sheets do not mention. In early childhood education, the object of planning is not a document displayed at a supervision visit. It is process quality: everyday interactions with adults, peers, materials and space, the most proximal driver of development, learning and well-being (OECD, 2021). A model does not observe today’s group. It produces a plausible text for a generic child. Inference, marked as such: the market for “AI for planning” inherits a category error from primary and secondary school — confusing preparing a resource with planning — and automates it in an age band where planning is, by definition, situated.

This paper does not recycle axes already treated in this series. The question is one of pedagogical category: what counts as everyday lesson planning when an early childhood centre “does AI.” The contributions are three: reconstruct the state of the art that separates the textual artefact generated a priori from observation, situated decision and in-the-act adjustment; examine three families of cases; and offer four tests for deciding when a kindergarten may claim that it plans, and not only that it produces a plan.

2. State of the art: from the textual artefact to planning as situated craft

Four strata that the “AI for planning” market usually mixes should be kept apart. The first is evidence that teachers use models to produce plans and resources, almost all of it measured in primary and secondary school, and almost all of it as preparation time or as quality of the document. The second is evidence of what early childhood education calls planning: group observation, emergent curriculum, spontaneous teaching and intentionality in the act. The third is the map of AI in early childhood education, which does not treat everyday planning as its object. The fourth is the rights and process-quality framework, which treats the 3–6-year-old as a subject of human oversight and the educator as the one who decides in the interaction.

In the artefact stratum, Roy, Poet, Staunton, Aston and Thomas (2024) publish the first large-scale randomised trial of ChatGPT in lesson preparation: 259 Year 7 and 8 science teachers in 68 English secondary schools, randomised by school. The group that used ChatGPT with a guide spent, in weeks six to ten, 69% of the time of the group asked not to use generative AI: 56.2 minutes per week versus 81.5, a saving of 25.3 minutes (31%). A blinded expert panel found no evidence of a quality difference in the resources. Status: empirical finding from secondary school, not ages 3–6. Kaufman, Woo, Eagan, Lee and Kassan (2025), with nationally representative U.S. data for 2023–2024, report that 25% of K–12 teachers used AI tools for instructional planning or teaching; use was nearly twice as high in English language arts and science (~40%) as among elementary teachers of all subjects or mathematics teachers (~20%). England lists lesson and curriculum planning among possible uses of generative AI and funded a lesson assistant (Department for Education, 2023/2025). Inference: the system that “does AI for planning” measures minutes of preparation and quality of a resource. It does not measure whether anyone observed this group or adjusted in the act.

In the early childhood planning stratum, NAEYC (2020, 2022) distinguishes observing and using that information in planning and in moment-to-moment interactions, and planning and implementing a curriculum for meaningful goals. Rönnlund (2025) conceptualizes spontaneous teaching in Swedish preschool as relational and balancing interaction: responding to the child’s focus of attention and expanding content without cancelling self-initiated play. Hjelmér, Rönnlund and Olausson (2025) saturate that picture with a one-year project and 50 preschool teachers: connecting the spontaneous with curricular goals was the hardest part. Speldewinde, Guarrella and Campbell (2025) observe, in Australian bush kinders, an emergent curriculum that combines child-centredness and teacher intentionality. Vitiello, Hutchins, Krissinger, Jirout and Scoville (2025) show, with 15 pre-k and kindergarten teachers and leaders, that the challenges of an emergent curriculum are not solved by a prior text. Government of Ireland (2024) and Siraj (2024) locate planning in a cycle of noticing and responding. SEP (2024) asks for narrative planning that starts from what was observed. Inference: in early childhood education, planning is a craft of seeing and deciding, not of drafting.

In the AI-in-ECE stratum, Su and Yang (2022) review 17 studies from 1995 to 2021. Chen (2024) maps 18 articles from 11 countries (2005–2023; 14 of them in 2020–2023) covering 15,081 children aged 2 to 8, and extracts four affordances: interactive tool; predicting or classifying the child’s conditions; adapting and personalizing; and a meaning axis. Ljungcrantz (2026) updates to 39 studies for 2020–2024, 20 of them in 2024, with a focus on ages 4–6. Status: field finding, not everyday planning. None of Chen’s affordances is “observe this group and decide in the act.” Inference: the 2024 surge does not authorize translating “more AI in the kindergarten” as “more planning.”

In the rights and process-quality stratum, the Recommendation on the Ethics of AI requires human oversight and particular attention when children are involved (UNESCO, 2021). Miao and Holmes (2023) set a threshold of 13 years for independent conversations with generative platforms and require ethical and pedagogical validation. UNICEF (2021) prioritizes the best interests of the child. Regulation (EU) 2024/1689 treats as high-risk, in Annex III, AI systems intended to determine access or admission to educational institutions, to assign persons, to evaluate learning or to determine the appropriate educational level (Unión Europea, 2024). The European Commission (2022) and the U.S. Department of Education (2023) agree that professional judgement must not be replaced. OECD (2021) anchors early childhood quality in everyday interactions; OECD (2023) documents digitalization in ECEC across 30 countries and jurisdictions; TALIS Starting Strong 2024 locates staff practices with children — not the production of documents — as a system lever (OECD, 2025). Macha, Hildebrandt, Wronski, Lonnemann and Urban (2024) recall that sustained shared thinking is a dimension of process quality, not a script. Inference: a five-year-old is not the user of a generated plan. The child is the subject of a decision an adult makes upon seeing them.

3. Review method

A critical narrative review was conducted, not a meta-analysis. The purpose was not to estimate a homogeneous effect size among an English secondary science trial, a Swedish preschool project and a composing platform, but to articulate an argument of pedagogical category with verified sources. Inclusion criteria: (a) 2021–2026, with NAEYC 2020 frameworks when they are the current edition; (b) everyday lesson planning, emergent curriculum, spontaneous teaching or AI production of plans, or AI in early childhood education ages 3–6; (c) relevance for kindergarten, preschool, CENDI or ages 3–6; (d) peer-reviewed journal, DOI or report from NAEYC, NCCA, SEP, UNESCO, OECD, UNICEF, the European Union, the European Commission, EEF/NFER, RAND, the Department for Education or ERIC; (e) verifiable DOI or publisher page. Axes already used in this series were excluded as central object, though some appear as a boundary. Sources on AI planning in teacher education are used as evidence of the artefact, not as an article on teacher education.

The search was executed on 26 August 2026 on DOI pages, Springer, Elsevier, Wiley, SAGE, MDPI, Taylor & Francis, OECD iLibrary, UNESDOC, UNICEF, EUR-Lex, ERIC, EEF, RAND, NAEYC, NCCA, SEP and publisher sites. Each source was verified against at least one of those pages. The analysis distinguished three enunciative statuses. Empirical finding: what was observed or measured in the sample. Conceptual or normative framework: what a framework or a regulation prescribes. Pedagogical inference: the translation to kindergartens and CENDI, marked as such.

4. Case 1. Producing a plan faster is not planning this group

Roy et al. (2024) publish the trial that best illustrates the leap the thesis rejects. In England, 34 schools (129 teachers) were randomly assigned to use ChatGPT with a guide to prepare Year 7 and 8 science lessons and resources, and 34 schools (130 teachers) not to use any generative AI. The trial lasted ten weeks in summer 2024. The primary result is time: 56.2 minutes per week versus 81.5, a saving of 25.3 minutes. The secondary result is the resource: the expert panel found no evidence that the quality of materials used by the two groups differed. Status of the evidence. Empirical finding of workload and of document quality in secondary school. It is not a kindergarten finding. It is not a finding of observation of a 3–6 group. It is not a finding of in-the-act adjustment. Pedagogical inference, marked as such: this is the gesture a CENDI copies when it “does AI for planning.” Time spent drafting a plan is taken as time spent planning, a text is produced faster, and planning is declared. What the trial authorizes is narrower: a model can shorten preparation of a secondary science resource, with no evidence that the resource is worse to a panel’s eye. It does not authorize saying that the kindergarten planned.

Karaman and Göksu (2024) saturate the artefact picture in primary school: 39 third-graders, 25 hours of mathematics over five weeks; the experimental group worked with ChatGPT plans. Within-sample achievement rose (d = 1.268); the between-group post-test difference was not significant. Status: third-grade finding, small N; the effect is not inflated. Yu, Lee and Kim (2025) bring the artefact closer to early childhood. At a Midwestern U.S. university, 44 teacher candidates (24 early childhood, 20 elementary) used ChatGPT to plan mathematics. Of 280 prompts, idea generation was 43.3% in the early childhood group. The finding that matters to this thesis is not the gain in confidence — an object of teacher education, not of this article — but the obstacle participants named: adapting the text to developmentally appropriate practice and to a specific classroom. Ko (2024), with 42 early childhood students in Korea, reports the same friction: age, developmental level, vague activity, hard-to-grasp goals. Gurl, Markinson and Artzt (2025) find, in a secondary mathematics microteaching, lessons that were teacher-centred, repetitive and with little knowledge of students’ needs. Status: production of a text, not a kindergarten in the act. Inference: the model does not fail because it is “not yet well trained on early childhood.” It fails because it cannot observe this group. A plausible plan for a generic four-year-old is not a plan for the eighteen children in this room on Tuesday at nine.

Kaufman et al. (2025) and the Department for Education (2023/2025) saturate the system picture: one-quarter of U.S. K–12 teachers report using AI to plan or teach; England lists lesson planning as a use and funds a lesson assistant. Chen (2024), Su and Yang (2022) and Ljungcrantz (2026) show that the “AI in ECE” field grows without treating everyday planning as its object. Inference: the gap is not of products. It is of category. The artefact is measured because the artefact is what a model knows how to produce.

5. Case 2. What the kindergarten does do when there is planning: observe, decide, adjust

Rönnlund (2025) publishes in the European Early Childhood Education Research Journal the study that best names, in this corpus, planning as craft in the act. From a one-year project in which Swedish preschool teachers documented and analysed spontaneous teaching, she conceptualizes that teaching as relational and balancing interaction. Relational: respond to the child’s focus of attention, expand content, integrate adult and child intentions dynamically. Balancing: negotiate curriculum goal orientations with interests and self-initiated play, without perceiving that balance as an insurmountable obstacle, in part because Swedish goals are broad and allow co-creation. Status of the evidence. Empirical finding of teachers’ conceptualization in preschool, not of AI. It is not a finding that a model reproduces that teaching. It is not, and this article does not turn it into, a free-play study as its axis: the object is spontaneous teaching as a form of planning in the act. Pedagogical inference, marked as such: this is the object an early childhood centre may properly call planning. It is seeing this group and deciding. A plan generated the night before does not respond to this morning’s focus of attention.

Hjelmér, Rönnlund and Olausson (2025) saturate the case with the same project: three researchers and 50 preschool teachers over a year. Analysis of their own and others’ spontaneous situations made interactions with children and the group’s interests visible. Connecting those situations with specific curricular goals was the most challenging aspect. The prerequisites were not a plan generator: they were time to analyse, collegial reflection and an analysis tool aligned with preschool values. Status: finding from N = 50 in Sweden, not generalizable to every CENDI; not an AI finding. Inference: if the hard part of everyday planning is articulating what has just happened with a goal, a model that writes the goal before anything happens does not solve the problem. It hides it.

Speldewinde, Guarrella and Campbell (2025) carry the same craft to Australian bush kinders: 32 interviews with 20 educators (2015–2023). The emergent science curriculum is not a prior text: it combines child-centredness and teacher intentionality, and gives children a voice to articulate understandings. Status: emergent-curriculum finding, not AI; play appears as a means, not as this article’s object. Vitiello et al. (2025), with 15 pre-k and kindergarten teachers and leaders, show that making an emergent curriculum work requires support, and that challenges worsen in public school. Macha et al. (2024) recall that sustained shared thinking extends in dialogue, not in a script. Siraj (2024) holds that the best curriculum is not taken off the shelf. Government of Ireland (2024) translates that into a cycle of noticing, nurturing, responding and reflecting, with spontaneous planning as a type. NAEYC (2020) asks that what is observed be used in planning and moment to moment. SEP (2024), under Agreement 07/08/23, asks for planning that is integral, open, flexible and narrative, starting from observing and recording. Inference: evidence of planning in early childhood is verified in the act — responding to the focus, articulating a goal to what has just occurred — not in a pdf generated at 10 p.m.

OECD (2021) offers the system contrast: process quality is everyday interactions, not planning documents. OECD (2025) locates staff practices with children as a lever. Inference: a kindergarten that exhibits ten generated plans and cannot narrate what it observed on Tuesday has not planned. It has archived.

6. Case 3. AI as the adult’s materials workshop, not as author of the day

Farrow, Farrow and Zhang (2025) publish in Education Sciences the case that best illustrates the third artefact when it is placed on the side of the document. Twenty-nine teachers in the southeastern United States used L4C, an AI-enhanced platform to support composing lesson plans in early childhood. After use, plans aligned 5.2 L4C principles versus 1.2 at baseline, t(28) = 8.2; quality practices in the plans rose from 33% to 53%, t(28) = 4.6, p < 0.001, d = 0.71. Fifty-two percent identified the scripted lesson as the most supportive feature. Status: finding of written-plan quality, not of an interaction trial in kindergarten. It is not a finding that AI “plans.” Pedagogical inference, marked as such: the script is exactly the artefact this thesis does not admit as planning. It can be a material. It cannot be the act. Yu et al. (2025) and Ko (2024) show why: the text does not know this group or today’s classroom. Gurl et al. (2025) show the bias: adult-centred, repetitive. Roy et al. (2024) measure the resource, not the act. Inference: if AI enters a kindergarten or a CENDI, the only place it does not contradict early childhood planning is the adult’s workshop — lists of materials, a draft discarded upon seeing the group — subjected to pedagogical validation. It does not enter as author of the day.

Miao and Holmes (2023) set the threshold of 13 years for independent conversations with generative platforms. A four-year-old is not the interlocutor of the plan or of the generated assembly. OECD (2023, 2025) locates digitalization in staff, materials and protection, not in a script read to the group. NAEYC (2020, 2022) reserves moment-to-moment decision to the educator. Government of Ireland (2024) reserves spontaneous planning to the adult who notices. SEP (2024) reserves narrative planning to whoever observed. Inference: a kindergarten that asks a model for “today’s assembly” has not planned the assembly. It has subcontracted the first act of the day to a system that was not at the door.

The rights framework closes the case. UNESCO (2021) requires human oversight. UNICEF (2021) requires development and well-being. The European Commission (2022) requires that professional judgement not be replaced. Annex III of Regulation (EU) 2024/1689 includes automated assessment and determination of educational level (Unión Europea, 2024). Inference: a system that, in a CENDI, generates the day’s sequence and treats it as a pedagogical decision is not a planning innovation. In that territory it is a high-risk practice if it determines level or assessment. Miao and Holmes’s (2023) threshold is not “met” by lowering the age of the plan’s user. It is respected by not turning the early childhood child into the addressee of a script no adult situated.

7. Inferential framework: four tests for claiming that there is planning, not an artefact

The framework that follows is this article’s pedagogical inference, anchored in the cases and in the verified instruments. It is not a new international standard. It distinguishes four tests. If a kindergarten, preschool, CENDI or infant school does not pass them, it cannot declare that a model-generated plan, a “ready-made” sequence or a prompt that produces assembly constitutes pedagogical planning.

7.1. Test of this group, not of the generic child. Yu et al. (2025) and Ko (2024) document the difficulty of adapting generated text to developmentally appropriate practice and to the specific classroom. Gurl et al. (2025) find lessons blind to students’ needs. NAEYC (2020) asks that observation inform planning of these children’s experiences. SEP (2024) asks for planning that addresses the needs and interests of the children of this context. Inference: evidence of planning is verified in traits of this group — who arrived, what they brought, what did not work yesterday — present in the decision. If the “evidence” that the centre plans is a prompt with the age and the week’s theme, the centre has generated a text for a generic child. It has not planned.

7.2. Test of prior or concurrent observation, not of the a priori text. Rönnlund (2025) and Hjelmér et al. (2025) locate spontaneous teaching in what is seen and analysed. Speldewinde et al. (2025) start from listening to children. Government of Ireland (2024) and Siraj (2024) start from noticing. NAEYC (2020) starts from observing. SEP (2024) starts from observing and recording. Roy et al. (2024), by contrast, measure the time to produce a resource before the class. Inference: a plan written the night before, without observation of this group, does not pass the test. It can be an input. Planning begins when someone looks.

7.3. Test of in-the-act adjustment, not of the closed sequence. Rönnlund (2025) describes the balance between goals and self-initiated play as the craft itself. Hjelmér et al. (2025) find that articulating the spontaneous with a goal is what is hard. Farrow et al. (2025) show that what is appreciated in a platform is often the script. Macha et al. (2024) recall that sustained shared thinking extends in dialogue, not in a script. Inference: if the assembly is read exactly as it came out of the model, there was no planning. There was execution. Early childhood planning includes throwing the plan away when the group is not there, or is somewhere else.

7.4. Test of the adult’s judgement, not of the model’s fluency. Miao and Holmes (2023) exclude children under 13 as independent interlocutors. The European Commission (2022) and the U.S. Department of Education (2023) require that professional judgement not be replaced. UNESCO (2021) requires human oversight. OECD (2021, 2025) locates quality in staff interactions. Inference: the 3–6-year-old is not the user of a generated plan. The child is the inhabitant of a morning an adult decides. If AI enters, it enters as the adult’s materials workshop, subjected to pedagogical validation. It does not enter as author of the assembly, the sequence or the day.

The framework admits generative AI as the adult’s workshop (Farrow et al., 2025; Yu et al., 2025, read against the thesis); cycles of noticing and responding (Government of Ireland, 2024; SEP, 2024); and spontaneous teaching and emergent curriculum as forms of planning (Rönnlund, 2025; Hjelmér et al., 2025; Speldewinde et al., 2025). It rejects declaring planning by a generated plan, a “ready-made” sequence or an assembly prompt (Roy et al., 2024; Yu et al., 2025; Ko, 2024; Farrow et al., 2025).

8. Discussion

Three tensions organize the discussion. The first is between producing a document and planning a group. It is a finding that 259 teachers in 68 secondary schools saved 31% of preparation time with no evidence of a quality difference in resources (Roy et al., 2024); that one-quarter of U.S. K–12 teachers report using AI to plan or teach (Kaufman et al., 2025); and that England lists that use (Department for Education, 2023/2025). It is a framework that early childhood planning starts from observation and is also exercised moment to moment (NAEYC, 2020; Government of Ireland, 2024; SEP, 2024). It is not a finding that producing a plan faster produces the spontaneous teaching of Rönnlund (2025) and Hjelmér et al. (2025) or the emergent curriculum of Speldewinde et al. (2025). The policy of the three artefacts measures what engineering knows how to measure and declares what only situated craft would authorize.

The second is between the generic child and this group. Yu et al. (2025) and Ko (2024) show the friction with developmentally appropriate practice and with the concrete classroom. Gurl et al. (2025) show an adult bias. Chen (2024) shows that the AI-in-ECE literature classifies and personalizes; it does not observe. Inference: insisting that “the plan is already there” while no one has seen this group is inverted pedagogy. NAEYC (2020) and SEP (2024) place observation at the start. The model writes first and, if at all, looks afterwards.

The third is between the adult’s workshop and the author of the day. Farrow et al. (2025) show that a platform can improve a written plan and that the script is perceived as support. Miao and Holmes (2023) exclude children under 13 as independent users. Inference: the only AI use that does not contradict early childhood planning is the one that remains on the adult’s side and is subjected to pedagogical validation, including the right to discard the text at the door. A list of materials can be planning if it starts from what was observed. A model that writes the assembly because the prompt asked for “Monday circle” is not.

9. Limits

This review is narrative. It does not apply PRISMA or estimate combined effects. Roy et al. (2024) are a secondary-school trial in England; transfer to ages 3–6 is inference. Karaman and Göksu (2024) are third grade, N = 39, with no between-group post-test difference. Yu et al. (2025) and Ko (2024) are initial teacher education: they are used as evidence of the artefact, not as an article on teacher education. Gurl et al. (2025) are secondary. Farrow et al. (2025) measure written-plan quality (N = 29), not process interactions; the script finding is read against the thesis. Rönnlund (2025) and Hjelmér et al. (2025) are Swedish preschool; N = 50 is not generalized to CENDI. Speldewinde et al. (2025) are bush kinders; play is cited as a means, not as object. Vitiello et al. (2025) are 15 interviews. Chen (2024), Su and Yang (2022) and Ljungcrantz (2026) map AI in ECE, not everyday planning. NAEYC, Aistear, SEP, Siraj, UNESCO, UNICEF, the OECD and the European Union are framework sources. Macha et al. (2024) are used for sustained shared thinking as process quality, not as a social-emotional learning axis. Kaufman et al. (2025) and the Department for Education (2023/2025) are school-system sources. No Latin American trials of AI and everyday lesson planning in kindergarten that measure observation and in-the-act adjustment against a generated plan were located. The inferences in section 7 are hypotheses of pedagogical category, not implementation evidence.

10. Conclusions

A model-generated lesson plan, a “ready-made” sequence or a prompt that produces the day’s assembly do not constitute pedagogical planning in an early childhood centre. Verified evidence does not authorize that declaration. Two hundred and fifty-nine science teachers in 68 secondary schools saved 31% of preparation time with ChatGPT, with no evidence of a quality difference in resources: that is production of a document, not planning of a 3–6 group (Roy et al., 2024). One-quarter of U.S. K–12 teachers report using AI to plan or teach (Kaufman et al., 2025). Early childhood teacher candidates stumble when adapting the text to developmentally appropriate practice and to the concrete classroom (Yu et al., 2025; Ko, 2024); suggested lessons come out adult-centred (Gurl et al., 2025); twenty-nine teachers improve a written plan and thank the script (Farrow et al., 2025). Eighteen studies covering 15,081 children aged 2 to 8 distinguish classifying and personalizing, not observing this group (Chen, 2024); reviews of 17 and 39 studies saturate the field without equating it to everyday planning (Su & Yang, 2022; Ljungcrantz, 2026). By contrast, when there is planning in early childhood, there is an adult who observes and decides: spontaneous teaching in Swedish preschool (Rönnlund, 2025; Hjelmér et al., 2025), emergent curriculum in bush kinders (Speldewinde et al., 2025), a cycle of noticing and responding (Government of Ireland, 2024; Siraj, 2024), narrative planning (SEP, 2024) and moment-to-moment interactions (NAEYC, 2020, 2022). Current law requires human oversight, the best interests of the child, an age threshold for independent conversations with generative platforms and, in the European Union, treating as high-risk the AI that determines access, admission, assignment, assessment or level (UNESCO, 2021; UNICEF, 2021; Miao & Holmes, 2023; Unión Europea, 2024; European Commission, 2022).

Where the sources do not measure a kindergarten, this article does not claim it. Where they measure a faster document or a higher-scoring script, it does not translate them into planning. Accompanying children aged three to six with everyday lesson planning is observing this group, deciding in situation and adjusting in the act. The rest is a textual artefact generated a priori. It is not planning, and it must not be presented as what it is not.

Laboratorio Editorial de NEXTECH.IA / Ingeniero Mitre.

References

  1. Chen, J. J. (2024). A scoping study on AI affordances in early childhood education: Mapping the global landscape, identifying research gaps, and charting future research directions. Journal of Artificial Intelligence Research, 81, 701–740. https://doi.org/10.1613/jair.1.16882
  2. Department for Education. (2023/2025). Generative artificial intelligence (AI) in education (updated 12 August 2025). GOV.UK. https://www.gov.uk/government/publications/generative-artificial-intelligence-in-education/generative-artificial-intelligence-ai-in-education
  3. European Commission. (2022). Ethical guidelines on the use of artificial intelligence (AI) and data in teaching and learning for educators. Publications Office of the European Union. https://doi.org/10.2766/153756
  4. Farrow, J., Farrow, M. J., & Zhang, C. (2025). Advancing translational science through AI-enhanced teacher learning for early language and composing. Education Sciences, 15(11), 1496. https://doi.org/10.3390/educsci15111496
  5. Government of Ireland. (2024). Aistear: The Early Childhood Curriculum Framework. National Council for Curriculum and Assessment. https://www.curriculumonline.ie/early-childhood/aistear-2024/
  6. Gurl, T. J., Markinson, M. P., & Artzt, A. F. (2025). Using ChatGPT as a lesson planning assistant with preservice secondary mathematics teachers. Digital Experiences in Mathematics Education, 11, 114–139. https://doi.org/10.1007/s40751-024-00162-9
  7. Hjelmér, C., Rönnlund, M., & Olausson, A. (2025). Analys av undervisning i spontant uppkomna situationer i förskolan. Educare, 2025(2), 75–101. https://doi.org/10.63310/edu.2025.2.54821
  8. Karaman, M. R., & Göksu, İ. (2024). Are lesson plans created by ChatGPT more effective? An experimental study. International Journal of Technology in Education, 7(1), 107–127. https://doi.org/10.46328/ijte.607
  9. Kaufman, J. H., Woo, A., Eagan, J., Lee, S., & Kassan, E. B. (2025). Uneven adoption of artificial intelligence tools among U.S. teachers and principals in the 2023–2024 school year (RR-A134-25). RAND Corporation. https://doi.org/10.7249/RRA134-25
  10. Ko, S. (2024). Exploring the process and experience of pre-service early childhood teacher creating a simulated instruction plan using ChatGPT. Korean Open Association for Early Childhood Education, 29(5), 315–337. https://doi.org/10.20437/KOAECE29-5-13
  11. Ljungcrantz, L. (2026). The interaction of AI and early childhood education. A state-of-the-art review 2020–2024. Early Childhood Education Journal, 54, 3565–3581. https://doi.org/10.1007/s10643-025-02079-3
  12. Macha, K., Hildebrandt, F., Wronski, C., Lonnemann, J., & Urban, M. (2024). Making it explicit – Sustained shared thinking dialogue as a way to explore children’s perspectives on quality in German early childhood education and care. British Educational Research Journal. https://doi.org/10.1002/berj.4054
  13. Miao, F., & Holmes, W. (2023). Guidance for generative AI in education and research. UNESCO. https://doi.org/10.54675/EWZM9535
  14. NAEYC. (2020). Developmentally Appropriate Practice (DAP) position statement. National Association for the Education of Young Children. https://www.naeyc.org/resources/position-statements/dap
  15. NAEYC. (2022). Developmentally Appropriate Practice in Early Childhood Programs Serving Children from Birth Through Age 8 (4th ed.). NAEYC. https://www.naeyc.org/resources/pubs/books/dap-fourth-edition
  16. OECD. (2021). Starting Strong VI: Supporting meaningful interactions in early childhood education and care. OECD Publishing. https://doi.org/10.1787/f47a06ae-en
  17. OECD. (2023). Empowering young children in the digital age (Starting Strong). OECD Publishing. https://doi.org/10.1787/50967622-en
  18. OECD. (2025). Results from TALIS Starting Strong 2024: Strengthening early childhood education and care. OECD Publishing. https://doi.org/10.1787/20af08c0-en
  19. Rönnlund, M. (2025). Preschool teachers’ conceptualizations of spontaneous teaching: Relational and balancing interactions. European Early Childhood Education Research Journal, 34(1), 5–18. https://doi.org/10.1080/1350293X.2025.2487594
  20. Roy, P., Poet, H., Staunton, R., Aston, K., & Thomas, D. (2024). ChatGPT in lesson preparation: A Teacher Choices trial. Evaluation report. Education Endowment Foundation / National Foundation for Educational Research. https://educationendowmentfoundation.org.uk/projects-and-evaluation/projects/choices-in-edtech-using-generative-ai-chatgpt-for-ks3-science-lesson-preparation-2024-teacher-choices-trial
  21. Secretaría de Educación Pública. (2023). Acuerdo número 07/08/23 por el que se determina el Currículo Nacional aplicable a la Educación Inicial: Programa Sintético de la Fase 1. Diario Oficial de la Federación. https://www.dof.gob.mx/nota_detalle.php?codigo=5698664&fecha=15/08/2023
  22. Secretaría de Educación Pública. (2024). La importancia de una planeación integral y narrativa en Educación Inicial. SEP. https://educacionbasica.sep.gob.mx/wp-content/uploads/2024/08/Inicial_-La-importancia-de-una-planeacion-integral-y-narrativa-en-Educacion-Inicial.pdf
  23. Siraj, I. (2024). An emergent and inquiry-based curriculum approach in early childhood. National Council for Curriculum and Assessment. https://ncca.ie/media/kcugefoz/an-emergent-and-inquiry-based-curriculum-approach-in-early-childhood.pdf
  24. Speldewinde, C., Guarrella, C., & Campbell, C. (2025). Listening to children in nature: Emergent curriculum in science teaching and learning in bush kinders. International Journal of Science Education, 47(15-16), 2180–2202. https://doi.org/10.1080/09500693.2024.2365459
  25. Su, J., & Yang, W. (2022). Artificial intelligence in early childhood education: A scoping review. Computers and Education: Artificial Intelligence, 3, 100049. https://doi.org/10.1016/j.caeai.2022.100049
  26. UNESCO. (2021). Recommendation on the ethics of artificial intelligence. UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000381137
  27. UNESCO. (2022). Tashkent Declaration and Commitments to Action for Transforming Early Childhood Care and Education. UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000384045
  28. UNICEF. (2021). Policy guidance on AI for children 2.0. UNICEF Innocenti. https://www.unicef.org/innocenti/reports/policy-guidance-ai-children
  29. Unión Europea. (2024). Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 (AI Act), Annex III. Official Journal of the European Union. https://eur-lex.europa.eu/eli/reg/2024/1689/oj
  30. U.S. Department of Education, Office of Educational Technology. (2023). Artificial intelligence and the future of teaching and learning: Insights and recommendations. U.S. Department of Education. https://www.ed.gov/sites/ed/files/documents/ai-report/ai-report.pdf
  31. Vitiello, V. E., Hutchins, N., Krissinger, A., Jirout, J., & Scoville, J. (2025). How early educators make emergent curricula work: A qualitative analysis with an eye toward scale. Early Education and Development, 37(1), 44–65. https://doi.org/10.1080/10409289.2025.2517401
  32. Yu, H. M., Lee, H., & Kim, S. H. (2025). Integrating ChatGPT in teacher education: Examining early childhood and elementary pre-service teachers’ experiences in AI-assisted math lesson planning. Journal of Research in Childhood Education, 40(1), 67–89. https://doi.org/10.1080/02568543.2025.2579668