1. Introduction and problem

In the 3–6 age range—nursery, preschool, CENDI, infant school—a package of four artifacts has been installed that purport to stand in for support for spatial reasoning, spatial language, or block play: a chatbot that “teaches spatial” or gives construction instructions to the child; a spatial-puzzles app that scores “spatial ability” or “spatial quotient”; a computer-vision model that classifies or scores the “complexity” or “quality” of a block construction without human mediation; and a generator that produces “spatial training” units from a prompt. All four are visible in coordination and allow the center to display that it “already does spatial with AI.” The leap—from generative instruction, score, complexity label, or unit to claiming support for spatial reasoning—is not authorized by evidence from AI mappings in ECE nor by block-play pedagogy when there is 2D/3D assembly with unit blocks, spatial vocabulary (location, dimension, shape, orientation), gestures, prolonged shared construction time, and human scaffolding.

The thesis of this article is restrictive. A chatbot that “teaches spatial” or gives construction instructions to the child, a spatial-puzzles app that scores “spatial ability” or “spatial quotient,” a computer-vision model that classifies or scores the “complexity” or “quality” of a block construction without human mediation, or a generator that produces “spatial training” units from a prompt do not constitute support for spatial reasoning in the early years. In early childhood, spatial reasoning develops in situated practice with blocks and open-ended concrete materials, spatial language between adult and child or among peers, gestures, prolonged construction time, and human scaffolding (modeling, gestural or linguistic spatial feedback); not an assembly score, an autonomous digital puzzle, nor an AI tutor that replaces the adult or peer. Bower et al. (2020) document spatial trainings (N = 187; age 3; MF/GF/SLF vs. control) with 2D/3D TOSA gains, especially in low-SES, and partial transfer to mathematics as a lateral outcome. Bower et al. (2022) compare digital app versus concrete materials versus control (N = 61): both improve 2D-TOSA versus control. Bower et al. (2025) extend with N = 331: concrete and digital intervention predicts spatial and math learning, with less robust transfer. Schmitt et al. (2024) create a coding system of spatial and quantitative language in block play (N = 24). Yang and Hu (2021) describe locations, deictics, dimensions, and shapes (N = 228). That authorizes asking what was measured: bot instruction, puzzle point, or complexity label, not the practice when a child places a block “on top,” turns a piece, and receives spatial feedback from an adult or peer.

The problem is aggravated by five category confusions. First: spatial reasoning / spatial language / block play is not STEM nor educational robotics—Berson et al. (2023) and Di Lieto et al. (2023) appear only as contrast, not as axis. Second: it is not numeracy as axis—mathematics can be a lateral outcome (Bower et al., 2020, 2025; Schmitt et al., 2024). Third: it is not artistic creativity—Aksoy and Aksoy (2022) show overlap, not construct identity. Fourth: it is not a “complexity” or “quality” score of construction by computer vision without human mediation. Fifth: it is not an autonomous digital puzzle that declares a “spatial quotient.” This work does not recycle STEM/robotics, numeracy, symbolic play, free play, artistic creativity, motor skills, executive functions, SEL, scientific inquiry, environmental education, literacy, orality, UDL, documentation, formative assessment, participation, planning, joint attention, adult–child interactions as a generic axis, or teacher training. The question is what counts as spatial reasoning when a center “does AI and spatial.”

There is, moreover, a coordination economy: tutor chatbot, puzzle score, complexity label, and printed unit fit on one slide; prolonged block talk with unit blocks, gestures, and adult–child spatial feedback do not. Ferrara, Hirsh-Pasek, Newcombe, Golinkoff and Lam (2011) anchor that guided play elevates spatial talk during block play. NAEYC (2022) and OECD (2021, 2023) demand meaningful interactions and subordinate digitalization. Pedagogical inference: support for spatial reasoning is not fulfilled by generating units or scoring assembly. The contributions are three: separate the craft from the artifact package; examine three families of cases—including BrickSmart as peripheral contrast when GenAI supports the adult/parent and does not replace the child (Liu et al., 2025; TRANSFER ages 6–8); and offer four tests to claim support for spatial reasoning, and not only chatbot, score, vision, or prompt.

2. State of the art: from situated block-play practice to the artifact on display

It is useful to separate four strata that the market of “AI for spatial reasoning in early childhood” usually mixes. The first is the construct of spatial reasoning / spatial language / block play for ages 3–6 as practice of 2D/3D assembly with unit blocks, vocabulary of location, dimension, shape, and orientation, gestures, and a prolonged shared-construction episode (Bower et al., 2020, 2022, 2025; Schmitt et al., 2024; Yang and Hu, 2021; Ferrara et al., 2011). The second is the pedagogical craft that cultivates it—open-ended concrete materials, TOSA / spatial language feedback / gesture feedback trainings, adult modeling, gestural or linguistic spatial feedback, peer scaffolding—(Bower et al., 2020, 2022, 2025; Ferrara et al., 2011; NAEYC, 2022; OECD, 2021, 2023). The third is evidence of AI affordances in ECE, child-centered GenAI, and systems that guide parents—without equating them to situated block play between adult and child nor to an autonomous tutor of the child—(Chen, 2024; Su and Yang, 2022; Su and Zhong, 2022; Ljungcrantz, 2026; Nikolopoulou, 2025; Liu et al., 2025, with TRANSFER marked). The fourth is the framework of rights, systems, and developmentally appropriate practice, which treats the 3–6 child as a subject of situated construction and spatial conversation, not as a vector of signals for a complexity-scoring pipeline nor as a passive receiver of chatbot instructions (UNESCO, 2021; Miao and Holmes, 2023; U.S. Department of Education, 2023).

In construct and craft, Bower et al. (2020) report an assembly intervention (N = 187; age 3; MF/GF/SLF vs. control): 2D/3D TOSA gains, stronger effect in low-SES, partial transfer to mathematics. Bower et al. (2022) compare digital app versus concrete materials versus control (N = 61): both improve 2D-TOSA versus control; they did not differ from each other. Bower et al. (2025)—N = 331; age 3—predict spatial and math learning; transfer to mathematics less robust. Schmitt et al. (2024) create a coding system of spatial and quantitative language in block play (N = 24). Yang and Hu (2021) describe locations, deictics, dimensions, and shapes (N = 228). Ferrara et al. (2011) anchor that guided play elevates block talk. Pedagogical inference: a chatbot does not replace “put it on top” said with a gesture; an adult or peer who scaffolds does.

In the artifact stratum, Chen (2024), Su and Yang (2022), Su and Zhong (2022), and Ljungcrantz (2026) map AI affordances in ECE without equating them to block play. Nikolopoulou (2025) balances GenAI under teacher mediation: caution framework. Liu et al. (2025) document BrickSmart: GenAI guides parents—does not replace the child—; ages 6–8 → TRANSFER marked; increase in spatial vocabulary. In category contrast, Berson et al. (2023) explore robot programming and CT in preschoolers (ages 3–4): robotics/CT ≠ block play as axis. Di Lieto et al. (2023) combine unplugged and robotics for CT/visuospatial skills: lateral contrast. Aksoy and Aksoy (2022) link block play and creativity: construct boundary. UNESCO (2021) demands human supervision; Miao and Holmes (2023) set pedagogical validation of GenAI; U.S. Department of Education (2023) demands that AI not replace professional judgment.

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 spatial-tutor chatbots, scoring puzzle apps, complexity classifiers, or unit generators, but to articulate a pedagogical category argument with verified sources. Inclusion criteria: (a) 2021–2026, with Ferrara et al. (2011) as block-talk anchor, and TRANSFER marked when the sample does not equate to ages 3–6 (e.g., BrickSmart 6–8); (b) spatial reasoning, spatial language, block play, spatial assembly, TOSA, gesture/spatial language feedback, or AI in ECE with artifact/craft relevance; (c) nursery, preschool, CENDI, or ages 3–6; (d) peer-reviewed journal, DOI, or NAEYC/UNESCO/OECD report; (e) verifiable DOI or publisher page. Excluded as central object were axes already used in this series—STEM/robotics, numeracy, symbolic play, free play, artistic creativity, motor skills, executive functions, SEL, scientific inquiry, environmental education, literacy, orality, UDL, documentation, formative assessment, participation, planning, joint attention, adult–child interactions as a generic axis, and teacher training. Robotics/CT only as contrast (Berson et al., 2023; Di Lieto et al., 2023). Numeracy only as a lateral outcome (Bower et al., 2020, 2025; Schmitt et al., 2024).

The search was executed on September 2, 2026 (slot 17:02 America/Mexico_City) on DOI pages, Crossref, Springer, Elsevier, APA PsycNet, Frontiers, ACM Digital Library, JAIR, OECD iLibrary, UNESDOC, NAEYC, and publisher sites. Each source was verified against at least one of those pages. Empirical finding, conceptual or normative framework, and pedagogical inference marked as such were distinguished. Priority was given to the distinction spatial reasoning / STEM-robotics / numeracy / creativity / CV score of block-construction complexity, and to the caution of not translating tutor chatbots, puzzle scores, or units-by-prompt into pedagogy of block play and mediated spatial language. N, d, r, AUC, and DOI were not invented: when an artifact (computer vision of construction complexity/quality; autonomous spatial-tutor chatbot) lacks a verified study in the corpus with scoring metrics in ages 3–6, it is discussed as a category ceiling supported by general AI mappings in ECE (Chen, 2024; Su and Yang, 2022; Ljungcrantz, 2026) and by the parent-mediated BrickSmart contrast (Liu et al., 2025), not as an invented empirical finding.

4. Case 1. “Spatial tutor” chatbot, scoring puzzle app, computer vision of construction complexity, or units-by-prompt do not constitute support for spatial reasoning

Chen (2024), Su and Yang (2022), and Ljungcrantz (2026) saturate the portrait of the artifact ceiling when AI in ECE is presented as if it were support for spatial reasoning. Chen (2024) maps global affordances of AI in early childhood education: empirical finding of scoping on emerging uses—tutoring, analytics, content generation—not empirical finding that a chatbot giving construction instructions cultivates block talk or mediated 2D/3D assembly. Su and Yang (2022) review the field of AI in ECE: empirical finding of synthesis on trends, not on spatial language of location, dimension, shape, or orientation during construction. Ljungcrantz (2026) reviews AI–ECE interaction 2020–2024: empirical finding of state of the art, not of situated block play. Pedagogical inference, marked as such: the gesture “the child received chatbot instructions = there was spatial reasoning” is a generative tutoring ceiling. Bower et al. (2020, 2022, 2025) and Ferrara et al. (2011) call for assembly with materials, gesture, spatial language, and guided play; a useful bot dialogue can coexist with absence of unit blocks, shared gestures, and prolonged construction time.

Nikolopoulou (2025) and Miao and Holmes (2023) name the risk of the “spatial training” units-by-prompt generator. Nikolopoulou (2025) balances promises and challenges of child-centered GenAI under teacher mediation: caution framework. Miao and Holmes (2023) demand pedagogical validation and age thresholds for generative AI. Status: normative and review framework, not trial of spatial units-by-prompt versus block play with adult–child spatial feedback. Pedagogical inference: producing a unit by prompt can be subordinate teacher preparation; support for spatial reasoning begins when there are blocks on the table, vocabulary of “on / under / beside / turn,” gestures that point, and an adult or peer who models and gives spatial feedback. Su and Zhong (2022) propose AI curriculum design in ECE as a future direction—AI literacy curriculum, not situated block play. Limit inference: an AI curriculum does not sign 2D/3D assembly.

The spatial-puzzles app that scores “spatial ability” or “spatial quotient” and the computer-vision model that classifies or scores the “complexity” or “quality” of a block construction without human mediation lack, in the verified corpus, trials with N, d, r, or AUC reported for complexity/quality scoring of constructions or algorithmic spatial quotient in preschool ages 3–6; figures are not invented. Chen (2024), Su and Yang (2022), and Ljungcrantz (2026) map analytics, tutoring, and classification as an AI-in-ECE trend, without equivalence to mediated block play. Bower et al. (2022) offer the decisive empirical contrast on digital versus concrete: both trainings improve 2D-TOSA versus control (N = 61), but the study does not authorize declaring that an app scoring “spatial quotient” constitutes the craft—it measures training with materials or app, not autonomous ability scoring as product. Liu et al. (2025)—TRANSFER 6–8—situate GenAI guiding parents in family block play: peripheral use, not child-directed tutor chatbot. Restrictive pedagogical inference: GenAI that supports the adult ≠ autonomous spatial tutor of the infant; mediated training app ≠ puzzle that declares quotient; teacher/parental empowerment ≠ CV score of tower complexity. UNESCO (2021) and U.S. Department of Education (2023) demand human supervision. The four artifacts share the same grammar of substitution: bot instruction replaces block talk; puzzle score replaces pedagogical observation of assembly; vision label replaces gestural/linguistic feedback; prompt replaces prolonged time with unit blocks.

It is useful to specify the ceiling without inventing effects. A chatbot can list steps; Yang and Hu (2021) and Schmitt et al. (2024) measure spatial language in human block play, not turns with a model. An app can sum points; Bower et al. (2020, 2025) measure TOSA and transfer, not quotient rankings. A classifier can label “high complexity”; Ferrara et al. (2011) anchor guided play, not pose detection. A generator can print a unit; Bower et al. (2020) document MF/GF/SLF trainings, not prompt output. Pedagogical inference: “high quotient” or “unit according to the model” is measurement or generation, not signed support.

5. Case 2. What the early childhood center does when there is support for spatial reasoning: block play, spatial language, gestures, and scaffolding—with BrickSmart as peripheral contrast

Bower et al. (2020) saturate the craft from controlled intervention (N = 187; age 3): spatial assembly trainings that incorporate materials, gesture (GF), and spatial language feedback (SLF) produce 2D/3D TOSA gains versus control, with greater benefit in low-SES and partial transfer to mathematics as a lateral outcome. Status: empirical finding. Pedagogical inference, marked as such: this is the object an early childhood center can call support for spatial reasoning. Spatial reasoning is not a puzzle score nor a complexity label: it is growing domain of assembling in 2D/3D with unit blocks, using vocabulary of location, dimension, shape, and orientation, gesturing during construction, sustaining the episode over time, and receiving human scaffolding—modeling, gestural or linguistic spatial feedback—from adult or peers. A chatbot that marks “next step” without shared blocks can celebrate a labeled dialogue and, at the same time, empty the construction table of spatial conversation. A complexity classifier can label “complex tower” in video and not have captured whether the adult protected time, modeled “turn the piece,” or invited a peer to describe where the block goes.

Bower et al. (2022) document that digital app and concrete materials improve 2D-TOSA versus control (N = 61) without differing from each other: digital can train when designed as training, not as autonomous scorer. Bower et al. (2025)—N = 331—confirm concrete and digital intervention; transfer to mathematics less robust. Schmitt et al. (2024) capture spatial and quantitative language (N = 24). Yang and Hu (2021) describe locations, deictics, dimensions, and shapes (N = 228). Ferrara et al. (2011) anchor guided play and block talk. NAEYC (2022) and OECD (2021, 2023) anchor meaningful interactions and subordinate digitalization.

The peripheral good-use contrast is BrickSmart. Liu et al. (2025) report GenAI to support spatial language learning in family block play: the system guides parents; it does not replace the child in construction nor converse as an autonomous tutor directed at the infant. Sample ages 6–8 → TRANSFER marked to the 3–6 cohort of this article. Status: empirical finding of increased spatial vocabulary under parental mediation assisted by GenAI. Pedagogical inference: subordinate digital—adult/parent who receives GenAI support to enrich block talk—can strengthen spatial language at the block table; it does not authorize declaring that a chatbot that “teaches spatial” to the child or an app that scores quotient constitutes the craft. Mediation—NAEYC (2022), OECD (2021)—is adult presence that organizes time, offers unit blocks, models spatial vocabulary, gestures, protects shared construction, and amplifies without replacing the child’s hand with an autonomous bot instruction. Pedagogical inference: pedagogical observation of spatial reasoning is professional reading of assembly and talk, not a log of complexity scores. A dashboard does not say “it is beside the red cube”; an adult or peer who scaffolds does.

6. Case 3. Spatial reasoning ≠ STEM/robotics, numeracy, creativity, or CV score of “block complexity”

The first category boundary is STEM/educational robotics. Programming a robot can coexist with lateral spatial skills; it does not, by itself, constitute block play of 2D/3D assembly with unit blocks, spatial language, and human scaffolding. Berson et al. (2023) explore robot programming and CT in preschoolers (ages 3–4): contrast—robotics/CT as a distinct empirical object. Di Lieto et al. (2023) combine unplugged and robotics for CT/visuospatial skills: lateral contrast. Pedagogical inference: this article forbids the equivalence “we have a robot = there is support for spatial reasoning via block play.” A spatial-tutor chatbot is not “digital STEM” either: it is generative tutoring.

The second boundary is numeracy as axis. Bower et al. (2020, 2025) and Schmitt et al. (2024) show mathematical transfer or overlap as a lateral outcome: empirical overlap, not construct identity. Pedagogical inference: a dashboard that rewards “block counting” does not convert numerical reinforcement into spatial reasoning. The third is artistic creativity: Aksoy and Aksoy (2022) show possible overlap, not equivalence. The fourth is the CV score of “complexity” or “quality” without human mediation: scoring the tower inverts pedagogy. UNESCO (2021), Miao and Holmes (2023), and U.S. Department of Education (2023) demand human supervision. Pedagogical inference: the only use of AI coherent with ages 3–6 remains on the adult’s side—BrickSmart guiding parents (Liu et al., 2025; TRANSFER) or subordinate teacher preparation—subject to pedagogical validation. It does not enter as autonomous scorer, tutor chatbot, quotient puzzle, or generator that replaces block talk.

7. Inferential framework: four tests to claim support for spatial reasoning, not an artifact

The following framework is pedagogical inference of this article, anchored in the cases and verified instruments. It is not a new international standard. It distinguishes four tests. If a nursery, preschool, CENDI, or infant school does not pass them, it cannot declare that a “spatial tutor” chatbot, a scoring puzzle app, computer vision of construction complexity/quality, or a spatial-training units-by-prompt generator constitute support for spatial reasoning.

7.1. Test of situated practice of 2D/3D assembly with unit blocks and spatial language (location, dimension, shape, orientation), not of the tutor chatbot nor the scoring puzzle. Bower et al. (2020, 2022, 2025), Yang and Hu (2021), and Schmitt et al. (2024) define the craft as observable assembly and spatial talk. Chen (2024), Su and Yang (2022), and Ljungcrantz (2026) map chatbots and content generation as affordances, not as block play. Pedagogical inference: evidence of support for spatial reasoning is verified in whether the child assembled, used spatial vocabulary, and gestured in shared construction. If the center’s “evidence” is a chatbot dialogue or a puzzle score, the center has done tutoring or gamification, not support for spatial block play.

7.2. Test of prolonged time, open-ended concrete materials, and scaffolding (gesture, spatial language feedback, modeling), not of the spatial-training unit by prompt. Ferrara et al. (2011), Bower et al. (2020), and NAEYC (2022) situate guided play, GF/SLF trainings, and developmentally appropriate practice. Miao and Holmes (2023) demand pedagogical validation of generative AI. Pedagogical inference: producing a unit by prompt does not demonstrate that there was protected time, unit blocks, or human spatial feedback. A script can exist; it does not sign the construction episode.

7.3. Test of human mediation—including the peripheral GenAI→adult/parent contrast—not of the complexity dashboard nor the autonomous tutor. OECD (2021, 2023) demand meaningful interactions and subordinate digitalization. Liu et al. (2025) anchor parent-mediated BrickSmart—TRANSFER 6–8. Pedagogical inference: “high complexity” or “spatial quotient” can raise a threshold without raising mediation quality. Pedagogical observation reads assembly and talk; the dashboard counts a label. GenAI that guides the adult is legitimate peripheral use; the chatbot that replaces the adult or peer in spatial conversation is not.

7.4. Test of category distinction and professional judgment, not of the product catalog. Spatial reasoning ≠ STEM/robotics, numeracy as axis, artistic creativity, or CV score of block complexity. Berson et al. (2023) and Di Lieto et al. (2023) prevent confusing robotics/CT with block play as axis. Bower et al. (2020, 2025) and Schmitt et al. (2024) prevent confusing lateral mathematical outcomes with construct identity. Aksoy and Aksoy (2022) prevent reducing block play to creativity. UNESCO (2021), Miao and Holmes (2023), U.S. Department of Education (2023), NAEYC (2022), and OECD (2021, 2023) demand human supervision, pedagogical validation, and not replacing professional judgment. Pedagogical inference: a center cannot treat the infant as a tower-complexity vector nor as exclusive interlocutor of a spatial tutor. Support for spatial reasoning is not fulfilled by scoring construction “quality” better. It is fulfilled by practicing 2D/3D assembly with unit blocks, spatial language, gestures, prolonged time, and human scaffolding.

The framework admits the digital when subordinated to teacher preparation, designed training, or validated parental mediation (Bower et al., 2022; Liu et al., 2025; Chen, 2024; Su and Yang, 2022; Ljungcrantz, 2026). It rejects declaring support for spatial reasoning by autonomous tutor chatbot, quotient-scoring puzzle app, unmediated computer vision of complexity, or unit generator that replaces block talk (Miao and Holmes, 2023; Nikolopoulou, 2025; UNESCO, 2021). The four tests are read together: passing only the first without mediation is not enough; passing only the fourth without situated practice is not either.

8. Discussion

Three tensions organize the discussion. The first is between displaying tutor chatbot, puzzle score, complexity vision, or generation and exercising support for spatial reasoning. It is empirical finding that assembly interventions (N = 187; N = 61; N = 331), coding of spatial language in block play (N = 24), the spatial-language inventory (N = 228), and the block-talk anchor sustain the situated craft (Bower et al., 2020, 2022, 2025; Schmitt et al., 2024; Yang and Hu, 2021; Ferrara et al., 2011). It is framework that AI in ECE grows without equivalence to mediated block play (Chen, 2024; Su and Yang, 2022; Ljungcrantz, 2026) and that BrickSmart inspires spatial vocabulary under parental mediation (Liu et al., 2025; TRANSFER). It is not empirical finding that the four artifacts produce the craft of the block table: they tutor, score, label, or generate what engineering delivers and declare what only mediated practice would authorize.

The second is between automated assessment of “spatial ability” or “construction complexity” and everyday pedagogy of block play. Chen (2024), Su and Yang (2022), and Ljungcrantz (2026) objectify analytics and classification affordances as a trend; they do not report in the corpus verified trials validating spatial-quotient scoring or CV of construction complexity/quality in preschool ages 3–6—hence metrics are not invented. Pedagogical inference: insisting that the nursery “already supports spatial reasoning” because a model classifies complexity or a chatbot gives instructions is inverted pedagogy. An algorithmic or conversational proxy is made to stand for assembly, spatial language, and gestures. The infant becomes a signal emitter for the classifier or a receiver of bot steps; the adult, a dashboard supervisor.

The third is between generated unit and integral craft at the block table. Miao and Holmes (2023) and Nikolopoulou (2025) set GenAI limits; Su and Zhong (2022) situate AI curriculum as a distinct construct. Pedagogical inference: selling a spatial-training unit by prompt as “spatial with AI” confuses textual product with situated practice. An autonomous chatbot because the system needs to direct steps does not replace the adult who places a block beside the child, says “it is under,” and gestures orientation. Bower et al. (2020, 2025) confirm lateral mathematical transfer without construct identity; Berson et al. (2023) confirm that robotics/CT is another category; UNESCO (2021), Miao and Holmes (2023), and U.S. Department of Education (2023) subordinate AI to professional judgment.

The four tests of section 7 read these tensions with operational criteria. The empirical craft contrast—N = 187; N = 61; N = 331; N = 24; N = 228; Ferrara et al.; mediated BrickSmart—defines the floor the four artifacts do not reach alone (Bower et al., 2020, 2022, 2025; Schmitt et al., 2024; Yang and Hu, 2021; Ferrara et al., 2011; Liu et al., 2025). Pedagogical inference: support erodes not by using AI, but by treating chatbot, score, vision, or generation as if they were the practice. Subordinate digital is legitimate—designed digital training (Bower et al., 2022); BrickSmart for parents (Liu et al., 2025)—; inverting the sequence is not. It is verified at the block table, with spatial language, gestures, and adult or peer who scaffolds (NAEYC, 2022; OECD, 2021, 2023).

9. Limits

This review is narrative. It does not apply PRISMA nor estimate primary combined effects. Bower et al. (2020, 2022, 2025) intervene N = 187 / 61 / 331 at age 3, not tutor chatbots nor autonomous quotient scoring. Schmitt et al. (2024) code N = 24; Yang and Hu (2021) describe N = 228; Ferrara et al. (2011) anchor block talk. Liu et al. (2025) document BrickSmart at ages 6–8: TRANSFER marked. Berson et al. (2023) and Di Lieto et al. (2023) are robotics/CT contrast. Aksoy and Aksoy (2022) measure creativity. Chen (2024), Su and Yang (2022), Su and Zhong (2022), Ljungcrantz (2026), and Nikolopoulou (2025) map AI in ECE, not complexity scoring. No verified trials of CV of construction quality nor of spatial-tutor chatbot in CENDI ages 3–6 were located; they are discussed as category ceiling. NAEYC, UNESCO, and OECD are framework sources. Section 7 inferences are pedagogical-category hypotheses, not implementation evidence.

10. Conclusions

A chatbot that “teaches spatial” or gives construction instructions to the child, a spatial-puzzles app that scores “spatial ability” or “spatial quotient,” a computer-vision model that classifies or scores the “complexity” or “quality” of a block construction without human mediation, or a generator that produces “spatial training” units from a prompt do not constitute support for spatial reasoning in an early childhood center. Chen (2024), Su and Yang (2022), and Ljungcrantz (2026) confirm AI affordances without equivalence to block play. Nikolopoulou (2025) and Miao and Holmes (2023) set GenAI limits. Liu et al. (2025) require reading BrickSmart as peripheral contrast—GenAI→parents ≠ tutor chatbot nor complexity score. When there is support, there is situated practice: trainings with gesture and spatial language feedback (Bower et al., 2020; N = 187); digital and concrete (Bower et al., 2022; N = 61); intervention N = 331 (Bower et al., 2025); coding of spatial language (Schmitt et al., 2024); spatial language (Yang and Hu, 2021); block talk (Ferrara et al., 2011); robotics/CT contrast (Berson et al., 2023; Di Lieto et al., 2023); creativity boundary (Aksoy and Aksoy, 2022); DAP and interactions (NAEYC, 2022; OECD, 2021, 2023). Spatial reasoning is distinguished from STEM/robotics, numeracy, creativity, and CV score. Guidelines demand human supervision (UNESCO, 2021; Miao and Holmes, 2023; U.S. Department of Education, 2023; Su and Zhong, 2022).

Where sources do not measure a nursery, this article does not claim it. Where they measure AI mappings, GenAI, or GenAI→parents, it does not translate them into pedagogical support for spatial reasoning. Accompanying three- to six-year-olds in spatial reasoning is exercising 2D/3D assembly with unit blocks, spatial vocabulary of location, dimension, shape, and orientation, gestures, prolonged shared construction time, and human scaffolding (modeling, gestural or linguistic spatial feedback). Everything else is tutor chatbot, quotient-scoring puzzle, complexity scoring, and training unit by prompt. It is not support for spatial reasoning in early childhood education, and it must not be presented as what it is not.

Laboratorio Editorial de NEXTECH.IA / Ingeniero Mitre.

References

  1. Aksoy, M., y Aksoy, A. B. (2022). An investigation on the effects of block play on the creativity of children. Early Child Development and Care, 193(1), 139–158. https://doi.org/10.1080/03004430.2022.2071266
  2. Berson, I. R., Berson, M. J., McKinnon, C., Aradhya, D., Alyaeesh, M., Luo, W., y Shapiro, B. R. (2023). An exploration of robot programming as a foundation for spatial reasoning and computational thinking in preschoolers’ guided play. Early Childhood Research Quarterly, 65, 57–67. https://doi.org/10.1016/j.ecresq.2023.05.015
  3. Bower, C., Zimmermann, L., Verdine, B., Toub, T. S., Islam, S., Foster, L., Evans, N., Odean, R., Cibischino, A., Pritulsky, C., Hirsh-Pasek, K., y Golinkoff, R. M. (2020). Piecing together the role of a spatial assembly intervention in preschoolers’ spatial and mathematics learning: Influences of gesture, spatial language, and socioeconomic status. Developmental Psychology, 56(4), 686–698. https://doi.org/10.1037/dev0000899
  4. Bower, C. A., Zimmermann, L., Verdine, B. N., Pritulsky, C., Golinkoff, R. M., y Hirsh-Pasek, K. (2022). Enhancing spatial skills of preschoolers from under-resourced backgrounds: A comparison of digital app vs. concrete materials. Developmental Science, 25(2), e13148. https://doi.org/10.1111/desc.13148
  5. Bower, C. A., Zimmermann, L., Verdine, B. N., Toub, T. S., Hirsh-Pasek, K., y Golinkoff, R. M. (2025). What does play have to do with it? A concrete and digital spatial intervention with 3-year-olds predicts spatial and math learning. Developmental Psychology. https://doi.org/10.1037/dev0001904
  6. Chen, J. J. (2024). A scoping study on AI affordances in early childhood education. Journal of Artificial Intelligence Research, 81, 701–740. https://doi.org/10.1613/jair.1.16882
  7. Di Lieto, M. C., et al. (2023). Combined unplugged and educational robotics training to promote computational thinking and cognitive abilities in preschoolers. Education Sciences, 13(9), 858. https://doi.org/10.3390/educsci13090858
  8. Ferrara, K., Hirsh-Pasek, K., Newcombe, N. S., Golinkoff, R. M., y Lam, W. S. (2011). Block talk: Spatial language during block play. Mind, Brain, and Education, 5(3), 143–151. https://doi.org/10.1111/j.1751-228X.2011.01122.x
  9. Liu, Y., Zha, S., Zhang, Y., Wang, Y., Zhang, Y., Xin, Q., Nie, L., Zhang, C., y Xu, Y. (2025). BrickSmart: Leveraging generative AI to support children’s spatial language learning in family block play. En Proceedings of the 2025 CHI Conference on Human Factors in Computing Systems (Article). ACM. https://doi.org/10.1145/3706598.3714212
  10. 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
  11. Miao, F., y Holmes, W. (2023). Guidance for generative AI in education and research. UNESCO. https://doi.org/10.54675/EWZM9535
  12. 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
  13. Nikolopoulou, K. (2025). Child-centered integration of generative AI in early learning: Balancing promises and challenges. AI, Brain and Child. https://doi.org/10.1007/s44436-025-00023-1
  14. OECD. (2021). Starting Strong VI: Supporting meaningful interactions in early childhood education and care. OECD Publishing. https://doi.org/10.1787/f47a06ae-en
  15. OECD. (2023). Empowering young children in the digital age (Starting Strong). OECD Publishing. https://doi.org/10.1787/50967622-en
  16. Schmitt, S. A., Duncan, R. J., Budreviciute, A., Purpura, D. J., y Eason, S. H. (2024). Capturing math language use during block play: Creation of the Spatial and Quantitative Mathematical Language Coding System. Journal of Numerical Cognition, 10, Article e11589. https://doi.org/10.5964/jnc.11589
  17. Su, J., y 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
  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, 100072. https://doi.org/10.1016/j.caeai.2022.100072
  19. 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
  20. UNESCO. (2021). Recommendation on the ethics of artificial intelligence. UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000381137
  21. Yang, X., y Hu, B. Y. (2021). Spatial language of young children during block play in kindergartens in urban China. Frontiers in Psychology, 12, Article 568638. https://doi.org/10.3389/fpsyg.2021.568638