1. Introduction and problem

In the 3-to-6-year span —kindergarten, preschool, CENDI, early childhood school— a package of five artifacts has taken hold that purport to count as support for mark-making / early drawing-writing / early strokes: a classifier/CNN/computer-vision system that «scores» strokes, shapes, Draw-A-Person, or fine motor maturity / prewriting readiness / drawing age / VMI auto-score scores as pedagogical assessment; a GenAI that generates mark-making lesson plans, prewriting worksheets, or copy-the-shape drills by prompt as a closed recipe; a dashboard of drawing minutes / stroke accuracy / shape completion for administrative surveillance; an app that replaces adult–child conversation about the meaning of marks with a correctness score; and a system that converts the portfolio into a ranking of developmental level from drawing without pedagogical mediation. The five allow centers to exhibit that they «already do mark-making with AI». The leap —from score, generated plan/worksheet, minutes metric, or maturational ranking to claiming support— is not authorized by AI-in-ECE mappings nor by mark-making pedagogy when there are intentional marks (scribble, strokes, shapes, human figure, prewriting), semiotic meaning and graphic agency, real materials (paper, pencil, chalk, paint; tablet subordinated), adult co-presence that observes, names, and converses —not only «copy the circle»—, and a process portfolio read with pedagogical mediation.

The thesis is restrictive: those five artifacts do not constitute support for mark-making / early drawing-writing in early childhood. In early childhood education, mark-making develops with real materials, graphic agency, and co-presence that interprets meaning and process (NAEYC, 2022; OECD, 2021, 2023); not a shape score, a GenAI of worksheets, or a minutes dashboard. AI may support adult preparation in a subordinated way; it must not replace the relationship or turn the child into a vector of stroke accuracy. Serpa-Andrade et al. (2021) contribute a support system for educators and therapists who rate prewriting acquisition in Cuenca (Ecuador), with a corpus on the order of three hundred drawings associated with Battelle: a finding of assisted rating, not of semiotic pedagogy. Hwang et al. (2025) automate fine-motor assessment via hand-drawn shapes at 36–72 months: a motor score via shapes, not conversation about meaning. Liu et al. (2026) develop and preliminarily evaluate CPAT for prewriting skills in Chinese kindergarteners (N = 143) with VMI/OA anchoring: assisted assessment, not the craft of graphic agency. Strikas et al. (2023), Beltzung et al. (2025), Jensen et al. (2023), and Alawad et al. (2026) saturate CNN and machine vision over kindergarten drawings, drawing age, or the human figure: they measure architecture or latent structure / estimated level, not co-presence that names marks. Tsai et al. (2024) and Lomurno et al. (2023) are read with marked transfer (VMI rehab; tablet-based dysgraphia risk screening). That authorizes asking what was measured: fine motor maturity, prewriting readiness, drawing age, VMI auto-score, stroke accuracy, or developmental ranking —not the practice when a child marks with intention and an adult converses about meaning.

The problem is aggravated by six category confusions. First: mark-making is not emergent literacy as the axis —there may be traces of emergent writing IN mark-making; the axis is the semiotic practice of marking—. Second: it is not motor skills as the axis —Hwang et al. (2025), Strikas et al. (2023), and Tsai et al. (2024) measure motor/VMI scores; they do not authorize reducing mark-making to a motor curriculum—. Third: it is not artistic creativity —there is drawing, but the axis is graphic agency and semiotic meaning—. Fourth: it is not generic formative assessment nor documentation —the portfolio matters as pedagogical reading—. Fifth: it is not clinical Draw-A-Person nor dysgraphia screening —Jensen et al. (2023), Alawad et al. (2026); Lomurno et al. (2023) with transfer—. Sixth: it is not dialogic reading, orality, joint attention, free play, UDL, STEM, numeracy, SEL, teacher education, sand/water/playdough/loose parts/outdoor/risky/block/spatial, inclusion, or privacy. Galbraith (2022) is a peripheral contrast of play pedagogies. The question is what counts as support for mark-making when a center «does AI and early drawing-writing».

There is a coordination economy: the five artifacts fit on a slide; the episode with paper or chalk, intentional marks, an adult who names, and a mediated portfolio does not. NAEYC (2022) and OECD (2021, 2023) require meaningful interactions and subordinated digitization. Inference: support is not fulfilled by scoring stroke accuracy or generating worksheets by prompt. Contributions: separating craft and artifacts; examining marked cases (Serpa-Andrade et al., 2021; Hwang et al., 2025; Liu et al., 2026; Strikas et al., 2023; Beltzung et al., 2025; Tsai et al., 2024; Jensen et al., 2023; Alawad et al., 2026; Lomurno et al., 2023; Galbraith, 2022 peripheral); and offering four tests of support.

2. State of the art: from situated mark-making practice to the artifact on display

It is useful to separate four strata that the «AI for mark-making in early childhood» market tends to mix. The first is the construct of mark-making / early drawing-writing / early strokes at ages 3 to 6 as a semiotic and relational practice: intentional marks, semiotic meaning, graphic agency, and adult–child conversation about what was marked (NAEYC, 2022; OECD, 2021, 2023). The second is the craft that cultivates it —real materials; an adult who observes, names, and converses, not only corrects form; a process portfolio with mediation; tablet subordinated— (Serpa-Andrade et al., 2021, read as rating ≠ semiotic pedagogy; NAEYC, 2022; OECD, 2021, 2023; Galbraith, 2022, peripheral contrast). The third is the evidence of AI affordances in ECE and of CNN/CV systems for scoring drawing, shapes, VMI, human figure, or prewriting —without equating them to situated mark-making— (Chen, 2024; Su and Yang, 2022; Su and Zhong, 2022; Ljungcrantz, 2026; Nikolopoulou, 2025; Hwang et al., 2025; Liu et al., 2026; Strikas et al., 2023; Beltzung et al., 2025; Tsai et al., 2024; Jensen et al., 2023; Alawad et al., 2026; Lomurno et al., 2023). The fourth is the framework of rights and developmentally appropriate practice, which treats the young child as a subject who marks with intention, not as a vector of stroke accuracy nor as the exclusive recipient of GenAI worksheets or developmental rankings (UNESCO, 2021; Miao and Holmes, 2023; U.S. Department of Education, 2023).

NAEYC (2022) and OECD (2021, 2023) anchor appropriate practice, meaningful interactions, and subordinated digitization: framework. The verified systems measure objects distinct from the craft. Serpa-Andrade et al. (2021) design and implement a support system for educators and therapists who rate prewriting skills: a finding of assisted rating, not of conversation about the meaning of marks. Hwang et al. (2025) contribute automatic fine-motor assessment via shapes (36–72 months). Liu et al. (2026) —CPAT; N = 143— contribute a preliminary evaluation of prewriting skills with VMI/OA anchoring. Strikas et al. (2023) compare CNN against Griffiths. Beltzung et al. (2025) —N = 193 (ages 2–10 plus adults)— drawing behaviour with deep learning: marked transfer. Jensen et al. (2023) and Alawad et al. (2026, Tebyan) address human figure / developmental levels without diagnosis. Tsai et al. (2024) apply deep learning to VMI in rehab: transfer ≠ pure 3–6 classroom. Lomurno et al. (2023) detect dysgraphia risk with a tablet: screening transfer. Inference: a prewriting readiness score does not name the meaning of a scribble; a child who marks and an adult who converses do. Chen (2024), Su and Yang (2022), Su and Zhong (2022), and Ljungcrantz (2026) map AI in ECE without equating it to situated mark-making. Nikolopoulou (2025) and Miao and Holmes (2023) set caution and GenAI thresholds for lesson plans/worksheets by prompt. UNESCO (2021) and U.S. Department of Education (2023) require human oversight and not replacing professional judgment. Galbraith (2022) is a peripheral contrast of play pedagogies.

3. Review method

A critical narrative review was conducted, not a primary meta-analysis. The purpose was not to estimate a homogeneous effect size of the five artifacts, but to articulate a pedagogical-category argument with verified sources. Inclusion criteria: (a) 2021–2026, with marked transfer when the sample is not equivalent to ages 3–6 in the classroom, is rehab, clinical, or dysgraphia screening, or the age range exceeds the span; (b) mark-making, early drawing-writing, prewriting, drawn shapes, human figure, VMI, CNN/CV of drawing, or AI in ECE with artifact/craft relevance; (c) kindergarten, preschool, CENDI, or ages 3–6, or explicit transfer; (d) peer-reviewed journal, DOI, or NAEYC/UNESCO/OECD report; (e) verifiable DOI or editorial page. Excluded as central objects were axes already used in this series —emergent literacy, orality, motor skills as axis, artistic creativity, generic formative assessment, documentation, joint attention, dialogic reading, free play, UDL, STEM, numeracy, SEL, teacher education, sand/water/playdough/loose parts/outdoor/risky/block/spatial, inclusion, and privacy—. Galbraith (2022) only as peripheral contrast. Tsai et al. (2024) = VMI rehab, transfer. Lomurno et al. (2023) = dysgraphia/tablet screening, transfer. Beltzung et al. (2025) = ages 2–10 plus adults, partial transfer. Hwang et al. (2025) and Strikas et al. (2023) are read for fine-motor forms/CNN, not as the already published motor-skills axis. Serpa-Andrade et al. (2021) and Liu et al. (2026) are read for prewriting rating/CPAT, not as emergent literacy or generic formative assessment as axes.

The search was executed on September 6, 2026 (slot 15:40 America/Mexico_City) on DOI pages, Crossref, Springer, Elsevier, Wiley, Taylor & Francis, MDPI, IEEE Xplore, Frontiers, WSEAS, JAIR, OECD iLibrary, UNESDOC, NAEYC, and editorial sites. Each source was verified against the fuentes.md and crossref_check.json corpus. Empirical finding, framework, and pedagogical inference were distinguished. No N, d, r, AUC, or DOI was invented: when an artifact lacks a verified study with stroke-accuracy dashboard metrics, GenAI of worksheets as a display package in Latin American CENDI, or an app that replaces conversation with correctness in a 3–6 classroom, it is discussed as a category ceiling supported by AI-in-ECE mappings (Chen, 2024; Su and Yang, 2022; Ljungcrantz, 2026) and by the verified scoring/CNN systems (Serpa-Andrade et al., 2021; Hwang et al., 2025; Liu et al., 2026; Strikas et al., 2023; Beltzung et al., 2025; Jensen et al., 2023; Alawad et al., 2026). The twenty-one sources of the verified corpus were used.

4. Case 1. CNN/CV score of strokes/shapes/DAP/fine-motor/prewriting readiness, GenAI of lesson plans/worksheets/copy-the-shape, minutes/stroke-accuracy dashboard, app that replaces conversation with a score, or portfolio→developmental-level ranking do not constitute support for mark-making

Chen (2024), Su and Yang (2022), and Ljungcrantz (2026) saturate the artifact ceiling when AI in ECE is presented as if it were support for mark-making. Chen (2024) maps global affordances: a scoping finding on emerging uses —tutoring, analytics, content generation—, not that a fine motor maturity score or a drawing age score cultivates graphic agency. Su and Yang (2022) review AI trends in ECE, not co-presence that names what was marked. Ljungcrantz (2026) reviews AI–ECE interaction 2020–2024: state of the art, not situated mark-making. Inference: «high stroke accuracy / prewriting readiness = there was support» is a shape-analytics ceiling. Serpa-Andrade et al. (2021), Hwang et al. (2025), and Liu et al. (2026) measure rating, fine motor via shapes, or CPAT; a score useful for administration can coexist with absence of co-presence that interprets meaning.

Nikolopoulou (2025) and Miao and Holmes (2023) name the risk of GenAI mark-making lesson plans / prewriting worksheets / copy-the-shape drills by prompt that close the episode: a framework of caution and age thresholds, not a trial versus an episode with paper and an adult who converses (NAEYC, 2022; OECD, 2021). Inference: producing a «copy the circle» drill by prompt may be subordinated preparation; support begins with real materials, agency to mark, and an adult who names without reducing to correct form. Su and Zhong (2022) propose an AI literacy curriculum, not situated mark-making.

The five artifacts lack, in the verified corpus, pedagogical-equivalence trials with the craft in Latin American CENDI ages 3–6; no d, r, or AUC of «support» is invented. Hwang et al. (2025), Strikas et al. (2023), and Beltzung et al. (2025) measure CNN/shapes/drawing age; Jensen et al. (2023) and Alawad et al. (2026) human figure / level; Tsai et al. (2024) and Lomurno et al. (2023) rehab/dysgraphia transfer; Galbraith (2022) peripheral contrast only when the adult sustains the craft. Restrictive inference: CNN ≠ adult who names; GenAI of worksheets ≠ open episode; dashboard ≠ observation; correctness app ≠ conversation; ranking ≠ portfolio with mediation. UNESCO (2021) and U.S. Department of Education (2023) require human oversight. The five share a grammar of substitution: the score speaks for the mark; the GenAI closes; the dashboard replaces observation; the app erases conversation; the ranking turns process into level without mediation. The display grammar inverts the sequence: first the artifact is activated; then «there is already mark-making with AI» is declared. Inference: exhibiting stroke accuracy or copy-the-shape drills resolves administrative coordination, not accompaniment of intentional marks. When the score speaks for the child and the worksheet speaks for the adult, the human-oversight threshold is unmet even if the dashboard looks «green».

5. Case 2. What the kindergarten does do when there is support for mark-making: semiotic-relational craft and what Serpa-Andrade, Hwang, Liu CPAT, Strikas, Beltzung, Tsai, Jensen, Tebyan, and Lomurno measure —with Galbraith as peripheral contrast

The floor of the craft is not a score: it is real materials (paper, pencil, chalk, paint; tablet subordinated), intentional marks, graphic agency, semiotic meaning, adult co-presence that observes, names, and converses —not only «copy the circle»—, and a process portfolio with mediation (NAEYC, 2022; OECD, 2021, 2023). Inference: there is support when that practice is protected; not when a fine motor maturity score, GenAI of worksheets, or developmental ranking is exhibited. The craft is recognized in the episode: the child marks with intention; the adult asks, names, and extends without closing on correctness; the tablet, if it appears, is a subordinated medium.

Serpa-Andrade, Pazos-Arias, Lopez-Nores, and Robles-Bykbaev (2021) saturate the Latin American anchor: a support system for educators and therapists who rate acquisition of prewriting skills; Cuenca (Ecuador); a corpus on the order of three hundred drawings of prewriting figures associated with Battelle. Status: empirical finding of assisted rating. Inference: they measure prewriting rating for educator/therapist; they do not equate to semiotic pedagogy of mark-making nor authorize presenting a shape CNN as assessment of the meaning of marks. Hwang et al. (2025) —36–72 months— contribute automatic fine-motor assessment via hand-drawn shape images. Marked distinction: cited for the shape-scoring artifact; this article is not converted into the already published motor-skills one. Inference: fine motor score ≠ conversation about what was marked. Liu et al. (2026) —CPAT; N = 143 Chinese kindergarteners; VMI/OA— contribute development and preliminary evaluation of computer-assisted assessment of Chinese prewriting skills. Inference: CPAT measures assisted prewriting skills; it does not sign graphic agency or a process portfolio with semiotic mediation.

Strikas et al. (2023) compare state-of-the-art CNN for assessing fine motor skills in Greek kindergarten drawings against Griffiths: CNN architecture ≠ craft of co-presence. Beltzung et al. (2025) —N = 193 (150 children ages 2–10 + 43 adults); France— use deep learning to study development of drawing behaviour: marked transfer; drawing age prediction does not replace the adult who names meaning at ages 3–6. Jensen et al. (2023) show rich latent structure in human figure drawings via human perception and machine vision: rich latent structure is not an administrative ranking without mediation. Alawad et al. (2026) —Tebyan; MobileNet/ResNet/EfficientNet— estimate developmental levels without claiming diagnosis: estimating level ≠ process portfolio with mediation; it does not authorize converting the portfolio into a ranking without an adult.

Tsai, Lee, and Huang (2024) apply deep learning to VMI in pediatric rehabilitation (Beery VMI; DenseNet): THERAPY/rehab; transfer ≠ pure 3–6 classroom; VMI auto-score does not authorize the five artifacts as kindergarten mark-making. Lomurno et al. (2023) —Play-Draw-Write; Procrustes; kindergarten→2nd grade— early dysgraphia risk with tablet: screening; transfer; ≠ mark-making pedagogy. Galbraith (2022) —play pedagogies in preservice teachers— is peripheral contrast, not a teacher-education axis. Coherent mediation (NAEYC, 2022; OECD, 2021, 2023): materials, observation of marks, conversation about meaning, and portfolio without a correctness script.

Taken together: there are systems that rate prewriting (Serpa-Andrade et al., 2021), score shapes/fine motor (Hwang et al., 2025; Strikas et al., 2023), assess prewriting (Liu et al., 2026), predict drawing age (Beltzung et al., 2025), analyze human figure (Jensen et al., 2023; Alawad et al., 2026), or screen dysgraphia/VMI with transfer (Tsai et al., 2024; Lomurno et al., 2023). None —by itself— signs graphic agency, semiotic meaning, and co-presence in the classroom. The technical evidence is real in its domain; «score = support for mark-making» is an illegitimate category inference.

6. Case 3. Mark-making ≠ emergent literacy as axis, motor skills as axis, artistic creativity, generic formative assessment, documentation, clinical Draw-A-Person, or dysgraphia screening

Six boundaries protect the axis. First: emergent literacy —there may be traces IN mark-making; the axis is the semiotic practice of marking—. Inference: «we work on literacy = there is mark-making» does not sign; nor is a prewriting readiness score literacy or pedagogical mark-making. Serpa-Andrade et al. (2021) and Liu et al. (2026) rate/assist prewriting; they do not authorize rewriting this article as literacy. Second: curricular motor skills —Hwang et al. (2025), Strikas et al. (2023), and Tsai et al. (2024) measure fine motor/VMI; they do not authorize reducing mark-making to a fine motor maturity score as if it were the craft—. Third: artistic creativity —there is drawing and human figure (Jensen et al., 2023; Alawad et al., 2026; Beltzung et al., 2025), but the axis is graphic agency and semiotic meaning—. Fourth: formative assessment and documentation —the portfolio matters; converting it into a developmental ranking without mediation (ceiling of Alawad et al., 2026, read restrictively) is neither mark-making nor those already published axes—. Fifth: clinical Draw-A-Person —Jensen et al. (2023) and Alawad et al. (2026) do not equate to clinical DAP nor to classroom pedagogy—. Sixth: dysgraphia screening —Lomurno et al. (2023) fix transfer; screening ≠ kindergarten mark-making—. UNESCO (2021), Miao and Holmes (2023), and U.S. Department of Education (2023) require human oversight. Inference: coherent AI at ages 3–6 remains on the adult side (subordinated preparation; Galbraith, 2022 peripheral), not as scorer, GenAI of copy-the-shape, dashboard, correctness app, or ranking without mediation.

These boundaries protect the article’s axis. Mark-making / early drawing-writing as a semiotic and relational practice may touch, laterally, literacy, fine motor skills, drawing, assessment, or documentation; it is not reducible to any of those already published axes. A center that «works on literacy», «works on fine motor», «does art», «assesses», or «documents» has not shown by that label alone that it cultivates intentional marks with graphic agency and co-presence that interprets meaning. Neither does a clinical DAP nor a dysgraphia screening transferred into classroom discourse. The distinction is one of pedagogical category, not a denial of lateral empirical overlaps.

7. Inferential framework: four tests for claiming that there is support for mark-making, not an artifact

The framework that follows is pedagogical inference of this article, 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 early childhood school does not pass them, it cannot declare that the five artifacts constitute support for mark-making / early drawing-writing.

7.1. Test of situated practice (real materials, intentional marks, graphic agency), not of the CNN/CV score or the correctness app. NAEYC (2022), OECD (2021, 2023), and the scoring cases (Serpa-Andrade et al., 2021; Hwang et al., 2025; Liu et al., 2026; Strikas et al., 2023; Beltzung et al., 2025; Jensen et al., 2023; Alawad et al., 2026) define the craft/scoring contrast. Chen (2024), Su and Yang (2022), and Ljungcrantz (2026) map analytics as affordance, not as mark-making. Inference: evidence of support is verified in whether the child marked with real materials and agency, and whether semiotic meaning was in play. If the «evidence» is fine motor maturity, drawing age, or a correctness score, the center did shape analytics, not support for mark-making.

7.2. Test of adult co-presence that observes, names, and converses —not only «copy the circle»—, not of GenAI lesson plans/worksheets/copy-the-shape. NAEYC (2022), OECD (2021), and Galbraith (2022, peripheral) situate practice and interactions. Miao and Holmes (2023) and Nikolopoulou (2025) require validation and mediation of GenAI. Inference: a script by prompt does not sign the mark-making episode.

7.3. Test of human mediation and the process portfolio, not of the minutes/stroke-accuracy dashboard or developmental ranking without mediation. OECD (2021, 2023) require meaningful interactions. Alawad et al. (2026) estimate levels without diagnosis; Tsai et al. (2024) and Lomurno et al. (2023) fix rehab/screening. Inference: high minutes or stroke accuracy do not by themselves raise co-presence. Subordinated adult preparation is legitimate; the scorer that turns the child into a shape vector is not.

7.4. Test of category distinction and professional judgment, not of the product catalog. Mark-making ≠ emergent literacy, motor skills, artistic creativity, formative assessment, documentation, clinical DAP, or dysgraphia screening as axes. Hwang et al. (2025), Strikas et al. (2023), Tsai et al. (2024), Lomurno et al. (2023), Jensen et al. (2023), and Alawad et al. (2026) block those confusions. UNESCO (2021), Miao and Holmes (2023), U.S. Department of Education (2023), NAEYC (2022), and OECD (2021, 2023) require human oversight and not replacing professional judgment. Inference: support is fulfilled with real materials, graphic agency, co-presence, and a mediated portfolio —not by scoring «form» or surveilling «minutes».

The framework admits digital subordinated to adult preparation (Galbraith, 2022 peripheral; Serpa-Andrade et al., 2021 as rating support, not substitution of conversation). It rejects declaring support via the five artifacts (Miao and Holmes, 2023; Nikolopoulou, 2025; UNESCO, 2021). The four tests are read together.

8. Discussion

Three tensions. First: exhibiting the five artifacts versus exercising support. Serpa-Andrade et al. (2021), Hwang et al. (2025), Liu et al. (2026; N = 143), Strikas et al. (2023), Beltzung et al. (2025; N = 193 with transfer), Tsai et al. (2024, rehab), Jensen et al. (2023), Alawad et al. (2026), Lomurno et al. (2023, dysgraphia), and Galbraith (2022 peripheral) sustain the contrast; Chen (2024), Su and Yang (2022), and Ljungcrantz (2026) map AI without equivalence to situated mark-making. Second: automated assessment versus semiotic-relational pedagogy —declaring support by stroke accuracy or shape completion is inverted pedagogy; no metrics of «pedagogical support» are invented—. Third: GenAI of worksheets / ranking / correctness app versus craft —Miao and Holmes (2023), Nikolopoulou (2025), NAEYC (2022), and OECD (2021, 2023)—: selling copy-the-shape or Tebyan-like ranking without mediation as «mark-making with AI» confuses product with practice. Serpa-Andrade et al. (2021) confirm rating in Cuenca without authorizing replacement of conversation; UNESCO (2021) and U.S. Department of Education (2023) subordinate AI to professional judgment.

The four tests in section 7 read these tensions. The empirical contrast —Cuenca; shapes 36–72 months; CPAT N = 143; CNN Greece; Beltzung N = 193 with transfer; VMI rehab; human figure; Tebyan; Lomurno screening; Galbraith peripheral— defines the floor that the artifacts do not reach alone when presented as mark-making pedagogy. Inference: support erodes when they are treated as if they were the practice. Subordinated digital is legitimate toward the adult (Serpa-Andrade et al., 2021 as rating support; Galbraith, 2022); inverting the sequence is not (NAEYC, 2022; OECD, 2021, 2023). If the center shows real materials, intentional marks, an adult who conversed about meaning and process, and a mediated portfolio, it may speak of support; if it only shows scores, GenAI worksheets, minutes, correctness, or ranking without mediation, it speaks of artifacts. Mixing both on a slide of «AI and mark-making» is the emptiness reconstructed here. Su and Zhong (2022): AI literacy curriculum ≠ situated early drawing-writing.

9. Limits

This review is narrative. It does not apply its own PRISMA nor estimate combined effects. Serpa-Andrade et al. (2021) contribute prewriting rating in Cuenca with a corpus on the order of three hundred: no d or r is invented; it is read as a tool, not as complete semiotic craft. Hwang et al. (2025) measure fine motor via shapes at 36–72 months: motor-construct limit. Liu et al. (2026) measure N = 143 CPAT: preliminary evaluation. Strikas et al. (2023) contribute a CNN comparison. Beltzung et al. (2025) measure N = 193 with ages 2–10 plus adults: partial transfer. Tsai et al. (2024) are VMI rehab: transfer. Jensen et al. (2023) and Alawad et al. (2026) delimit human figure/machine vision/level estimation. Lomurno et al. (2023) are dysgraphia/tablet screening: transfer. Galbraith (2022) is peripheral contrast, not an axis. Chen (2024), Su and Yang (2022), Su and Zhong (2022), Ljungcrantz (2026), and Nikolopoulou (2025) map AI in ECE, not stroke-accuracy dashboards or GenAI of worksheets as a display package in Latin American CENDI. No verified trials of the five artifacts as a single «mark-making with AI» package in a 3–6 classroom were located; they are discussed as a category ceiling. NAEYC, UNESCO, and OECD are framework sources. The inferences in section 7 are category hypotheses, not implementation evidence.

10. Conclusions

The five artifacts —CNN/CV of strokes/shapes/DAP or fine motor maturity / prewriting readiness / drawing age / VMI scores; GenAI of lesson plans/worksheets/copy-the-shape; minutes/stroke-accuracy dashboard; correctness app; developmental ranking without mediation— do not constitute support for mark-making / early drawing-writing in early childhood education. Chen (2024), Su and Yang (2022), and Ljungcrantz (2026) confirm affordances without equivalence to situated mark-making. Nikolopoulou (2025) and Miao and Holmes (2023) set GenAI limits. Serpa-Andrade et al. (2021), Hwang et al. (2025), Liu et al. (2026), Strikas et al. (2023), Beltzung et al. (2025), Tsai et al. (2024), Jensen et al. (2023), Alawad et al. (2026), and Lomurno et al. (2023) require reading scoring/CNN/VMI/human figure/dysgraphia for what they measure —with transfers— and for what they do not equate to semiotic pedagogy. Galbraith (2022) is peripheral contrast: adult ≠ scorer or GenAI of worksheets. When there is support there are real materials, intentional marks, graphic agency, co-presence that converses, and a mediated portfolio (NAEYC, 2022; OECD, 2021, 2023). Mark-making is distinguished from emergent literacy, motor skills, artistic creativity, formative assessment, documentation, clinical DAP, and dysgraphia screening as axes. The guidelines require human oversight (UNESCO, 2021; Miao and Holmes, 2023; U.S. Department of Education, 2023; Su and Zhong, 2022).

Where the sources do not measure a kindergarten, this article does not claim it. Where they measure AI mappings, prewriting rating, shape CNN, CPAT, drawing age, VMI rehab, human figure, Tebyan, or dysgraphia screening, it does not translate them into support via a shape score or a worksheet by prompt. Accompanying three-to-six-year-old girls and boys in mark-making / early drawing-writing is to exercise real materials (paper, pencil, chalk, paint; tablet subordinated), graphic agency, adult co-presence that interprets meaning and process —not only «copy the circle»— and a process portfolio with mediation. The rest is CNN/CV score, GenAI of worksheets/copy-the-shape, minutes/stroke-accuracy dashboard, correctness app, and developmental ranking. It is not support for mark-making in early childhood education, and it must not be presented as what it is not.

Editorial Laboratory of NEXTECH.IA / Ingeniero Mitre.

References

  1. Alawad, W. M., Alquayid, A., Alghofaily, S., Alharbi, S., y Alorainy, D. (2026). Tebyan: An AI-powered system for estimating developmental levels from children’s human figure drawings. Health Informatics Journal, 32(2). https://doi.org/10.1177/14604582261460248
  2. Beltzung, B., Pelé, M., Martinet, L., Maitre, E., Falck, J., y Sueur, C. (2025). Using deep learning predictions to study the development of drawing behaviour in children. Displays, 90, Article 103166. https://doi.org/10.1016/j.displa.2025.103166
  3. 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
  4. Galbraith, J. (2022). “A prescription for play”: Developing early childhood preservice teachers’ pedagogies of play. Journal of Early Childhood Teacher Education, 43(3), 474–494. https://doi.org/10.1080/10901027.2022.2054035
  5. Hwang, N.-H., Chen, S.-S., Pai, T.-W., Ko, M. H.-J., Yu, Y.-L., y Chen, H.-J. (2025). Automatic assessment of fine motor development in children through hand-drawn shape images. Pediatrics & Neonatology, 66(6), 606–612. https://doi.org/10.1016/j.pedneo.2025.04.001
  6. Jensen, C. A., Sumanthiran, D., Kirkorian, H. L., Travers, B. G., Rosengren, K. S., y Rogers, T. T. (2023). Human perception and machine vision reveal rich latent structure in human figure drawings. Frontiers in Psychology, 14, Article 1029808. https://doi.org/10.3389/fpsyg.2023.1029808
  7. Liu, Z., Wu, D., Peng, S., Dong, C., Huo, Y., Zhang, Y., Zhang, J., Zhong, W., Liu, D., y Chen, J. (2026). Development and preliminary evaluation of a computer-assisted assessment tool for Chinese prewriting skills in preschoolers. Frontiers in Psychology, 17, Article 1793395. https://doi.org/10.3389/fpsyg.2026.1793395
  8. Ljungcrantz, L. (2026). The interaction of AI and early childhood education. A state-of-the-art review 2020–2024. Early Childhood Education Journal, 54(5), 3565–3581. https://doi.org/10.1007/s10643-025-02079-3
  9. Lomurno, E., Dui, L. G., Gatto, M., Bollettino, M., Matteucci, M., y Ferrante, S. (2023). Deep learning and Procrustes analysis for early dysgraphia risk detection with a tablet application. Life, 13(3), Article 598. https://doi.org/10.3390/life13030598
  10. Miao, F., y Holmes, W. (2023). Guidance for generative AI in education and research. UNESCO. https://doi.org/10.54675/EWZM9535
  11. NAEYC. (2022). Developmentally appropriate practice in early childhood programs serving children from birth through age 8 (4.ª ed.). NAEYC. https://www.naeyc.org/resources/pubs/books/dap-fourth-edition
  12. Nikolopoulou, K. (2025). Child-centered integration of generative AI in early learning: Balancing promises and challenges. AI, Brain and Child, 1(1). https://doi.org/10.1007/s44436-025-00023-1
  13. OECD. (2021). Starting Strong VI: Supporting meaningful interactions in early childhood education and care. OECD Publishing. https://doi.org/10.1787/f47a06ae-en
  14. OECD. (2023). Empowering young children in the digital age (Starting Strong). OECD Publishing. https://doi.org/10.1787/50967622-en
  15. Serpa-Andrade, L. J., Pazos-Arias, J. J., Lopez-Nores, M., y Robles-Bykbaev, V. E. (2021). Design, implementation and evaluation of a support system for educators and therapists to rate the acquisition of pre-writing skills. IEEE Access, 9, 77920–77929. https://doi.org/10.1109/access.2021.3083496
  16. Strikas, K., Papaioannou, N., Stamatopoulos, I., Angeioplastis, A., Tsimpiris, A., Varsamis, D., y Giagazoglou, P. (2023). State-of-the-art CNN architectures for assessing fine motor skills: A comparative study. WSEAS Transactions on Advances in Engineering Education, 20, 44–51. https://doi.org/10.37394/232010.2023.20.7
  17. Su, J., y Yang, W. (2022). Artificial intelligence in early childhood education: A scoping review. Computers and Education: Artificial Intelligence, 3, Article 100049. https://doi.org/10.1016/j.caeai.2022.100049
  18. Su, J., y Zhong, Y. (2022). Artificial Intelligence (AI) in early childhood education: Curriculum design and future directions. Computers and Education: Artificial Intelligence, 3, Article 100072. https://doi.org/10.1016/j.caeai.2022.100072
  19. Tsai, Y.-T., Lee, J.-S., y Huang, C.-Y. (2024). Research on applying deep learning to visual–motor integration assessment systems in pediatric rehabilitation medicine. Algorithms, 17(9), Article 413. https://doi.org/10.3390/a17090413
  20. 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
  21. UNESCO. (2021). Recommendation on the ethics of artificial intelligence. UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000381137