1. Introduction and problem

In the 3-to-6 age band —kindergarten, preschool, CENDI, infant school— a package of four artefacts has been installed that claim to count as in-service teacher professional development. The first is a prompt course: a short sequence that teaches how to “talk to” a generative model and issues a “AI teacher” certificate. The second is a webinar: a synchronous hour, often massive, archived as evidence that staff “updated themselves”. The third is a badge or micro-credential: a digital seal that attests to having completed a module. The fourth is AI-generated microlearning: content pills that a model produces, personalises or recommends, and that the adult consumes on a phone between shifts. All four are visible and cheap in coordination time. They allow a setting to exhibit that it “already does TPD with artificial intelligence”. The enunciative leap is enormous: one moves from consuming content —or from accrediting its consumption— to asserting that professional development has occurred. That leap is authorised neither by the evidence on in-service TPD nor by what early childhood classrooms measure when staff change practice.

The thesis of this article is restrictive. A prompt course, a webinar, a badge or a piece of AI-generated microlearning does not constitute teacher professional development in early childhood education. Effective TPD is inquiry situated in practice with peers and accompaniment; not consumption of automated content. Sims, Fletcher-Wood, O’Mara-Eves, Cottingham, Stansfield, Goodrich, Van Herwegen, and Anders (2025), with 104 randomised trials, show that TPD that moves learner achievement combines mechanisms that produce insight, motivate change, develop techniques and, decisively, embed them in practice. A webinar embeds nothing. A badge does not observe the room. A prompt does not accompany. NAEYC requires that the professional practice of the early childhood educator be continuous, reflective and anchored in situated judgement with children, not in a module detached from the room (NAEYC, 2022). An algorithm that generates a pill and delivers it to the adult is not educating. It is administering content that the craft did not interpret.

The problem is aggravated by a craft reason that product sheets do not mention. In-service TPD is not the initial preparation of someone who does not yet have a group. It is not an event, nor a catalogue of “AI competencies”, nor an accumulation of hours. What is professional is not a property of the artefact, but the use of evidence from one’s own practice to modify what is done with the group (Sims et al., 2025; OECD, 2021). Confusing the product —course, webinar, badge, pill— with the process is the category error this paper names. TALIS Starting Strong 2024 documents that TPD for ECEC staff still predominantly takes the form of courses and seminars —more than 60% of staff— while job-embedded training and collaborative learning are less frequent, although they predict child-centred practices (OECD, 2025). Inference, marked as such: the market for “AI for early childhood TPD” inherits that distribution and automates it.

This paper does not recycle the axes already treated in this series. It is not an article on pre-service initial training. The question is one of pedagogical category: what counts as in-service TPD when an early childhood setting “does AI”. The contributions are three: to reconstruct the state of the art that separates situated inquiry from content consumption; to examine three families of cases; and to offer four tests for deciding when a kindergarten may claim that it does TPD, and not merely that it accredits attendance, badges or prompts.

2. State of the art: from embedded TPD to content that is consumed

It is useful to separate four strata that the market for “AI for early childhood TPD” tends to mix. The first is TPD as a process of practice change: insight, motivation, technique and embedding on the job (Sims et al., 2025). The second is evidence on formats in ECEC: coaching as a TPD tool (Yang, Huang, Su, Zhu, Hsieh, and Li, 2022), the course versus coaching (Downer et al., 2024), workshops plus on-site mentoring (Huang, Siraj, and Melhuish, 2024) and collaborative learning among educators (Liu, Hedges, and Cooper, 2024). The third is content consumption —micro-courses, micro-credentials, microlearning, generative AI for the adult— and its empirical ceiling (Clayback et al., 2024; Pölzl-Stefanec and Geißler, 2022; Fan et al., 2024; Doss, Wolfe, Tekkumru-Kisa, Christianson, Ziegler, and Kaufman, 2024; Zhang and Wang, 2025). The fourth is the rights and systems frame, which treats 3–6 staff as subjects of process-quality TPD, not as users of a catalogue (OECD, 2021, 2025; NAEYC, 2022; UNESCO, 2021; Miao and Holmes, 2023; European Commission, 2022).

In the process stratum, Sims et al. (2025) offer the most recent explanatory account: 104 randomised trials; an average TPD effect on standardised tests of 0.05 standard deviations; and a framework of fourteen mechanisms grouped as insight, motivation, techniques and embedding. Programmes with (almost) all mechanisms approach 0.17; those that incorporate none approach zero. Status: finding from a meta-analysis of in-service TPD, not isolated to ages 3–6. Inference, marked as such: a prompt course, a webinar or a badge covers, at best, the insight pole —and often not even that, if the content is not verified in the room. They do not cover embedding. Without embedding there is no TPD that the meta-analysis can call effective.

In the ECEC-format stratum, Yang, Huang, Su, Zhu, Hsieh, and Li (2022) review 33 randomised or quasi-experimental trials of coaching for early childhood teachers: positive effects on adults’ knowledge and instruction and on a range of child outcomes. Downer et al. (2024) separate, among 496 early childhood teachers, a fourteen-week course on interactions from individual coaching focused on the same interactions: the course moves knowledge and detection; coaching moves observed instructional support and children’s engagement with the adult; the combination adds no extra benefit. Huang, Siraj, and Melhuish (2024) adapt a TPD programme in Shenzhen with face-to-face workshops and on-site mentoring: 95 classrooms, 202 teachers, 547 children aged 3 to 5; effects on classroom quality (ECERS-E, ES = 0.81; SSTEW, ES = 0.72) and on children’s literacy (ES = 0.16). Liu, Hedges, and Cooper (2024) group, in a content analysis, the structural, motivational and sustainability conditions of collaborative learning for early childhood teachers. Status: finding that TPD that changes the room is accompaniment and inquiry, not an event. Inference: the market sells the event.

In the consumption stratum, Clayback, LoCasale-Crouch, and Alamos (2024) pilot an online, self-paced micro-course with 25 in-service educators: high satisfaction and self-reported change in knowledge and comfort; no classroom observation. Pölzl-Stefanec and Geißler (2022) evaluate online microlearning for Austrian educators: participation may help if workplace conditions are considered; they do not equate consumption with process TPD. Fan, Wei, Lewis, and Watts (2024), with 147 educators and 24 interviews in one U.S. state, find more perceived benefits among those who design or assess the micro-credential than among those who pursue it. Doss et al. (2024) evaluate a state STEM micro-credential project: progress below target and no detectable effects on mathematics and science tests. Zhang and Wang (2025), with 143 in-service teachers and 2,053 contributions in a course mediated by a Chinese generative model, find more information analysis and critical thinking, and less social interaction and collaboration. Status: finding of a consumption ceiling. Inference: AI can speed the analysis of a text. It does not, by itself, produce the inquiry with peers that early childhood TPD requires.

In the systems stratum, OECD (2021) anchors ECEC quality in process interactions and treats TPD as a lever of that quality, not as an accumulation of hours. OECD (2025) reports that online TPD is associated, in several systems, with lower perceived impact than in-person or hybrid formats, and that coaching reaches fewer than 40% of staff in most systems. OECD (2023) locates ECEC digitalisation in staff and protection, not in a feed of pills. NAEYC (2022) requires reflective professionalism. UNESCO (2021) requires human oversight. Miao and Holmes (2023) require pedagogical validation. The European Commission (2022) and the U.S. Department of Education (2023) agree on not substituting professional judgement. Miao and Cukurova (2024) are cited only as a limit: a list of AI competencies —including an “AI for professional learning” axis— is not TPD, and is not used here as a central object. Inference: a kindergarten does not demonstrate TPD by exhibiting that its staff “already have the AI competency”. It demonstrates it when room practice changes with peers and accompaniment.

3. Review method

A critical narrative review was conducted, not a meta-analysis. The purpose was not to estimate a homogeneous effect size among a micro-course, a coaching trial and a community of educators, but to articulate an argument of pedagogical category with verified sources. Inclusion criteria: (a) 2021–2026; (b) in-service TPD —not pre-service initial training— or AI applied to the adult’s professional learning; (c) relevance to kindergarten, preschool, CENDI or ages 3–6, or transfer explicitly marked when the sample is not early childhood; (d) peer-reviewed journal, DOI, or report from NAEYC, UNESCO, OECD, the European Commission or a department of education; (e) verifiable DOI or publisher page. Axes already used in this series —including initial training— were excluded as central objects, though some appear as a limit.

The search was executed on 27 August 2026 on DOI pages, Springer, Elsevier, Wiley, SAGE, Frontiers, APA, OECD iLibrary, UNESDOC, ERIC, NAEYC, RAND 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 regulation prescribes. Pedagogical inference: the translation to kindergartens and CENDI centres, marked as such.

4. Case 1. Consuming content —a prompt course, a webinar, a badge, microlearning— is not doing TPD

Clayback, LoCasale-Crouch, and Alamos (2024) publish in Early Childhood Research Quarterly the pilot that best names the fourth artefact when it is presented as TPD. Twenty-five in-service educators complete, in spring 2022, a brief, online, self-paced micro-course. Engagement and satisfaction are high. Participants report changes in knowledge and comfort with behaviour-support practices. Status of the evidence. Empirical finding of consumption: a micro-course can please and can shift what the educator says she knows. It is not a finding of observed change in the room, nor that the format constitutes TPD. The authors do not claim that the micro-course replaces coaching. They present it as a scalable path precisely because coaching “demands time and resources”. Pedagogical inference, marked as such: this is the gesture a kindergarten copies when it “does TPD with AI microlearning”. A workplace need is condensed into pills, satisfaction is measured, and development is declared. What exists is accredited consumption. TPD, in Sims et al. (2025) and NAEYC (2022), requires embedding in practice. A micro-course does not observe it. This article does not convert the study into a socio-emotional-climate paper: the object retained is the format —brief content, self-paced, self-report— and its ceiling.

Pölzl-Stefanec and Geißler (2022) saturate the portrait of microlearning among early childhood educators. In Austria they evaluate an online micro-step course and conclude that participation “might” improve professional capacities if the conditions of the post are considered. Status: finding of design and conditions of use, not that microlearning is process TPD. Inference: the conditional clause is the datum. Fan, Wei, Lewis, and Watts (2024) add the badge: 147 educators and 24 interviews; those who design or assess the seal report more benefits than those who pursue it. Status: finding of perceptions, not isolated early childhood. Inference, marked as such: the badge rewards the record of the issuer more than the craft of the earner. Doss et al. (2024), in a RAND evaluation of STEM micro-credentials in Louisiana, find participation below target and no detectable effect on mathematics and science tests. Status: K–12 finding, not ages 3–6. Inference: if the seal does not move achievement where standardised tests exist, a kindergarten cannot treat it as TPD merely by issuing it.

Zhang and Wang (2025) publish the case that best names generative AI when it is presented as TPD. One hundred forty-three in-service teachers take a programme mediated by a Chinese generative model; 2,053 contributions are analysed. The model increases information analysis and critical thinking. It performs less well on social interaction and collaboration. Status: finding from in-service teachers, not from kindergarten. Inference, marked as such: this is the operation the market sells as a “prompt course”. The model speeds analysis. It does not produce inquiry with peers. The deficit is not a detail: it is the object the thesis claims. Su and Yang (2022) and Chen (2024) map AI in ECE (17 and 18 studies; 15,081 children aged 2 to 8). Ljungcrantz (2026) updates 2020–2024: 39 studies, focus on ages 4–6. Status: field finding of AI in the children’s room, not of adult TPD. Inference: the 2024 boom does not authorise translating “more AI in the kindergarten” as “more staff TPD”.

TALIS Starting Strong 2024 closes the case with systems evidence. More than 60% of staff take part in courses and seminars. Job-embedded training and collaborative learning are less common. In six pre-primary entities and two under-3 settings, those who took online courses perceive lower overall TPD impact than those who took in-person or hybrid formats (OECD, 2025). Inference: a mass webinar and asynchronous microlearning occupy exactly the cell that staff themselves report as lower-impact. Being “AI-powered” does not change the cell.

5. Case 2. What the kindergarten does when there is TPD: accompaniment in practice, not a course that is accredited

Downer, Doyle, Pianta, Burchinal, Field, Hamre, LoCasale-Crouch, Howes, LaParo, and Scott-Little (2024) publish in Early Education and Development the trial that best names, in this corpus, the category difference between a course and accompaniment. Four hundred ninety-six early childhood teachers, recruited in community preschools and Head Start, are randomly assigned in two phases. Phase I is a fourteen-week course on effective interactions. Phase II, with re-randomisation, is individual coaching focused on the same interactions. Knowledge, skill in detecting interactions, observed interactions, beliefs about intentional teaching and children’s engagement are evaluated. The course increases knowledge and detection, and instructional support observed the following year. Coaching also raises instructional support, and children in those classrooms show more positive engagement with the adult. The combination adds no extra benefit for teachers or for the group. Status of the evidence. Empirical finding of process in a dense experimental N: a course moves what is known and detected; accompaniment moves what is done with the group and what the group does with the adult. It is not a finding about AI, nor that a webinar reproduces coaching. The context is U.S. preschool; it is not generalised to a Latin American CENDI. Pedagogical inference, marked as such: this is the object an early childhood centre may properly call TPD. It is inquiry into one’s own interaction with accompaniment. A prompt course does not observe the room. A badge does not re-randomise the craft. Microlearning does not measure children’s engagement. And the combination datum is the one the market does not cite: adding the course to coaching does not add. The event does not potentiate accompaniment. It duplicates it without gain.

Yang, Huang, Su, Zhu, Hsieh, and Li (2022) saturate the portrait with 33 trials. Coaching, as TPD, moves adults’ knowledge and instruction and a range of child outcomes. Status: synthesis finding, not a single programme and not AI. Inference: the format that early childhood evidence supports is not the webinar. It is reiterated accompaniment of a practice. Huang, Siraj, and Melhuish (2024) add a cluster trial in Shenzhen: 24 kindergartens (12 and 12), 95 classrooms, 202 teachers, 547 children aged 3 to 5. The TPD combines an intensive day, six half-day sessions and fortnightly on-site mentoring. It is not used here as a leadership article: the object retained is in-service TPD with workshops plus accompaniment in the room. Effects on ECERS-E (ES = 0.81) and SSTEW (ES = 0.72); children’s literacy ES = 0.16 and executive function ES = 0.27; no effect on numeracy or socio-emotional scores. Status: finding of a workshop-plus-mentoring package, not of microlearning. Inference: when early childhood TPD moves process quality, there is an adult who returns to the room with another adult who observes and talks. Sims et al. (2025) read the same craft at another scale: without embedding, the effect approaches zero. OECD (2025) reports that coaching reaches fewer than 40% of staff in most systems. Inference: the kindergarten that “already did the AI webinar” has not covered that absence. It has disguised it.

6. Case 3. Inquiry with peers is not a personalised feed

Guerra, Figueroa Céspedes, Zamora, and González (2024) publish in Early Years the Latin American case that best illustrates, in this corpus, situated inquiry with peers. They study two professional learning communities of early childhood teachers in Chile: 27 participants (12 and 15), with non-participant observation, interviews and focus groups, and a grounded-theory analysis. Each community highlights different aspects —pedagogical practice, self-care, trust and institutional validation— and the educators name the community as a space of complex reflective practice and collaborative conceptual learning, not merely of information acquisition. Status of the evidence. Qualitative empirical finding of in-service TPD in Latin American early childhood. It is not a finding about AI, nor that a platform reproduces that community. The N is not generalised to a national system. Pedagogical inference, marked as such: this is the second pole of the craft the thesis defends. TPD is not only one-to-one coaching. It is also inquiry with peers into one’s own practice. A microlearning feed, even if “personalised” by a model, is not a community. There is no trust, no validation, no contradiction among peers.

Yang, Zhang, Zhou, and Ji (2023) saturate the portrait with 28 preschool teachers in China, also with grounded theory on interviews and diaries. They extract five community features: common vision, a read–practice–share flow, continuous reflection, distributed leadership and peer and organisational support. The flow the teachers name is not “consume–accredit–archive”. It is read, take to the room and come back to share. Status: finding of perceptions of a situated PLC, not a trial. Inference: the “practice” in the flow is the category criterion. A badge does not practise. A webinar does not return to the room. Liu, Hedges, and Cooper (2024) group, for early childhood teachers, structural features (how collaborative learning is organised), motivational features (what deepens it) and sustainability features (what maintains the change). Status: finding from a content analysis of literature, not a new trial. Inference: sustainability is the pole that microlearning claims to solve —“it fits in the pocket, it is done between shifts”— and that the evidence locates elsewhere: in protected time, in trust and in the return to practice with others. OECD (2025) confirms, at systems scale, that embedded collaborative learning predicts child-centred and adaptive practices. Inference: a kindergarten does not “collaborate” because its staff receive the same pill email. It collaborates when one person’s practice becomes material for another’s, with accompaniment.

The distribution of roles is confirmed by contrast. Zhang and Wang (2025) show that generative AI performs less well precisely where this case locates the craft: social interaction and collaboration. Downer et al. (2024) show that the course does not replace accompaniment. Clayback et al. (2024) show a micro-course that moves self-report. Fan et al. (2024) show a seal more satisfactory for the issuer. Inference: AI enters early childhood TPD, if at all, on the side of the adult who inquires with peers and of whoever accompanies that inquiry —as a workshop for analysing a record, as support for a joint-observation protocol— subject to pedagogical validation (Miao and Holmes, 2023). It does not enter as an interlocutor that replaces the peer, nor as an issuer of a badge. Miao and Cukurova (2024) are cited here only as a limit: the “AI for professional learning” axis of a competency framework does not authorise treating a prompt course as TPD. A list does not embed.

7. Inferential framework: four tests for claiming that there is TPD, not accredited content

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 prompt course, a webinar, a badge or AI-generated microlearning constitutes in-service TPD.

7.1. Test of embedding in practice, not of the accredited event. Sims et al. (2025) locate effective TPD in combinations that reach embedding of techniques on the job. Downer et al. (2024) observe that the course moves knowledge and coaching moves the room. OECD (2025) finds lower perceived impact for online TPD in several systems. Clayback et al. (2024) measure satisfaction and self-report. Inference: TPD evidence is verified in the use of one’s own practice —observed, discussed, retried— not in an attendance log, a badge or a prompt certificate. If the “evidence” that the setting does TPD is an archived webinar, the setting has done an event, not the craft.

7.2. Test of accompaniment and of peers, not of the personalised feed. Yang et al. (2022) synthesise coaching as early childhood TPD. Huang, Siraj, and Melhuish (2024) combine workshops and on-site mentoring. Guerra et al. (2024) and Yang, Zhang, Zhou, and Ji (2023) describe communities that read, practise and share. Zhang and Wang (2025) find that generative AI does not sustain collaboration. Inference: the 3–6 adult is not the user of a model that “continuously trains her”. She is the subject of a craft exercised with other adults and with whoever accompanies. If AI enters, it enters as support for that inquiry, not as a substitute peer or automatic coach.

7.3. Test of change in the room and in the group, not of declared knowledge. Downer et al. (2024) measure children’s engagement only in the coaching arm. Huang, Siraj, and Melhuish (2024) measure ECERS-E, SSTEW and literacy. Clayback et al. (2024) do not observe the room. Doss et al. (2024) detect no effect on tests. Inference: an increase in “comfort with AI” or in prompt knowledge is not early childhood TPD. TPD is played out in the next move of the activity with the group —changing the material, regrouping, returning a prompt, sustaining a shared observation— not in a seal that updates.

7.4. Test of rights and of professional judgement, not of a competency catalogue. UNESCO (2021) requires human oversight. Miao and Holmes (2023) require pedagogical validation. The European Commission (2022) and the U.S. Department of Education (2023) require not substituting the teacher. NAEYC (2022) requires reflective professionalism. Miao and Cukurova (2024) are read as a limit, not as an object: an AI competency framework is not, by itself, a TPD programme. Inference: an early childhood setting cannot treat the educator as a consumer of an automated “AI updating” itinerary. TPD is not fulfilled by accrediting better. It is fulfilled by inquiring into practice with peers and accompaniment.

The framework admits coaching and on-site mentoring (Downer et al., 2024; Yang et al., 2022; Huang, Siraj, and Melhuish, 2024); inquiry communities (Guerra et al., 2024; Yang, Zhang, Zhou, and Ji, 2023; Liu et al., 2024); embedding mechanisms (Sims et al., 2025); and the predictor of child-centred practices of embedded collaborative TPD (OECD, 2025). It rejects declaring TPD by a prompt course, a webinar, a badge or AI-generated microlearning (Clayback et al., 2024; Pölzl-Stefanec and Geißler, 2022; Fan et al., 2024; Doss et al., 2024; Zhang and Wang, 2025; OECD, 2025).

8. Discussion

Three tensions organise the discussion. The first is between consuming and exercising. It is a finding that ECEC staff TPD is delivered, predominantly, as courses and seminars, and that the online format is associated in several systems with lower perceived impact (OECD, 2025); that a micro-course moves self-report (Clayback et al., 2024); that a seal satisfies the issuer more (Fan et al., 2024); and that a generative model speeds analysis and not collaboration (Zhang and Wang, 2025). It is a framework that effective TPD combines insight with embedding (Sims et al., 2025) and that early childhood professionalism is reflective and situated (NAEYC, 2022). It is not a finding that a prompt course, a webinar, a badge or microlearning produce the craft that Downer et al. (2024), Guerra et al. (2024) and Yang, Zhang, Zhou, and Ji (2023) observe or hear. The policy of the four artefacts measures what the platform knows how to accredit and declares what only situated inquiry would authorise.

The second is between the event and accompaniment. Downer et al. (2024) deliver the datum the market does not use: the course does not potentiate coaching. Yang et al. (2022) and Huang, Siraj, and Melhuish (2024) show packages that return to the room. OECD (2025) shows that coaching does not reach most staff. Inference: insisting that “the AI webinar has already been done” while no one observes the interaction is inverted pedagogy. The certificate blames the educator for what the system did not accompany.

The third is between the peer and the model. Guerra et al. (2024) show educators whose learning is named as reflective and collaborative. Yang, Zhang, Zhou, and Ji (2023) name the read–practice–share flow. Zhang and Wang (2025) measure generative AI’s collaboration deficit. Miao and Holmes (2023) require pedagogical validation. Inference: the only use of AI that does not contradict in-service TPD at ages 3–6 is the one that remains on the side of the adult who inquires with peers and of whoever accompanies, and is subjected to validation. A model that helps a pair analyse a record from their own room may serve TPD. A model that generates microlearning because the system needs a badge does not.

9. Limits

This review is narrative. It does not apply PRISMA or estimate combined effects. Sims et al. (2025) cover in-service TPD not isolated to ages 3–6. Downer et al. (2024) are U.S. preschool. Huang, Siraj, and Melhuish (2024) are Shenzhen and workshops plus mentoring, not AI; they are not used as a leadership article. Yang et al. (2022) synthesise coaching, not generative AI. Clayback et al. (2024) are an N = 25 of self-report. Fan et al. (2024) and Doss et al. (2024) do not isolate kindergarten; Doss is STEM. Zhang and Wang (2025) are not CENDI. Guerra et al. (2024) and Yang, Zhang, Zhou, and Ji (2023) are small-N qualitative studies. Chen (2024), Su and Yang (2022) and Ljungcrantz (2026) map AI in ECE, not adult TPD. Miao and Cukurova (2024) are cited only as a limit. NAEYC, UNESCO, OECD and the 2021–2023 guidelines are framework. No Latin American trials of in-service TPD and AI in kindergarten that measure situated inquiry against microlearning were located. The inferences in section 7 are hypotheses of pedagogical category, not implementation evidence.

10. Conclusions

A prompt course, a webinar, a badge or a piece of AI-generated microlearning does not constitute in-service teacher professional development in an early childhood setting. The verified evidence does not authorise that declaration. One hundred four trials locate effective TPD in mechanisms that reach embedding of practice, not in the event (Sims et al., 2025). More than 60% of ECEC staff still receive courses and seminars, coaching does not reach most, and the online format is associated, in several systems, with lower perceived impact (OECD, 2025). A micro-course moves self-report (Clayback et al., 2024). A seal satisfies the issuer more (Fan et al., 2024) and, in a state project, does not move achievement (Doss et al., 2024). A generative model speeds analysis and not collaboration (Zhang and Wang, 2025). Eighteen and thirty-nine reviews saturate the growth of AI in ECE without equating it to adult TPD (Chen, 2024; Su and Yang, 2022; Ljungcrantz, 2026). By contrast, when there is TPD in early childhood, there is accompaniment and inquiry: 496 teachers in whom the course neither replaces nor potentiates coaching (Downer et al., 2024), 33 coaching trials (Yang et al., 2022), a package of workshops plus on-site mentoring that moves process quality (Huang, Siraj, and Melhuish, 2024), 27 Chilean educators who name the community as reflective practice (Guerra et al., 2024) and 28 preschool teachers who name the read–practice–share flow (Yang, Zhang, Zhou, and Ji, 2023). Current law and guidance require human oversight, pedagogical validation and not substituting professional judgement; an AI competency framework is not, by itself, TPD (UNESCO, 2021; Miao and Holmes, 2023; European Commission, 2022; U.S. Department of Education, 2023; Miao and Cukurova, 2024).

Where the sources do not measure a kindergarten, this article does not assert it. Where they measure a course, a webinar, a badge or microlearning, it does not translate them into TPD. Accompanying educators of children aged three to six with professional development is to exercise inquiry situated in practice, with peers and accompaniment. Anything else is consumption of automated content. It is not TPD, and it should not be presented as what it is not.

Editorial Laboratory of 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. Clayback, K. A., LoCasale-Crouch, J., & Alamos, P. (2024). A pilot study of a micro-course to promote positive teaching practices and prevent exclusionary discipline in early childhood. Early Childhood Research Quarterly, 67, 182–190. https://doi.org/10.1016/j.ecresq.2023.12.010
  3. Doss, C. J., Wolfe, R. L., Tekkumru-Kisa, M., Christianson, K., Ziegler, M. D., & Kaufman, J. H. (2024). The role of micro-credentials in strengthening STEM teaching and learning: An evaluation of the Louisiana STEM Micro-Credentials Project (RR-A2486-2). RAND Corporation. https://doi.org/10.7249/RRA2486-2
  4. Downer, J. T., Doyle, N. B., Pianta, R. C., Burchinal, M., Field, S., Hamre, B. K., LoCasale-Crouch, J., Howes, C., LaParo, K., & Scott-Little, C. (2024). Coaching and coursework focused on teacher–child interactions during language/literacy instruction: Effects on teacher outcomes and children’s classroom engagement. Early Education and Development, 35(5), 1032–1062. https://doi.org/10.1080/10409289.2024.2303604
  5. 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
  6. Fan, X., Wei, X., Lewis, A., & Watts, J. (2024). Micro-credentialing as an approach to educator professional learning: Voices from developers, pursuers, and assessors. Teaching and Teacher Education, 148, 104700. https://doi.org/10.1016/j.tate.2024.104700
  7. Guerra, P., Figueroa Céspedes, I., Zamora, P., & González, M. P. (2024). An exploratory study of Chilean early childhood teachers’ professional learning communities. Early Years, 44(3-4), 585–599. https://doi.org/10.1080/09575146.2023.2198164
  8. Huang, R., Siraj, I., & Melhuish, E. (2024). Promoting effective teaching and learning through a professional development program: A randomized controlled trial. Journal of Educational Psychology, 116(8), 1352–1367. https://doi.org/10.1037/edu0000851
  9. Liu, M., Hedges, H., & Cooper, M. (2024). Effective collaborative learning for early childhood teachers: Structural, motivational and sustainable features. Professional Development in Education, 50(2), 420–438. https://doi.org/10.1080/19415257.2023.2235578
  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., & Cukurova, M. (2024). AI competency framework for teachers. UNESCO. https://doi.org/10.54675/ZJTE2084
  12. Miao, F., & Holmes, W. (2023). Guidance for generative AI in education and research. UNESCO. https://doi.org/10.54675/EWZM9535
  13. 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
  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. OECD. (2025). Results from TALIS Starting Strong 2024: Strengthening early childhood education and care. OECD Publishing. https://doi.org/10.1787/20af08c0-en
  17. Pölzl-Stefanec, E., & Geißler, C. (2022). “Micro-steps” on the route to successful online professional development for Austrian early childhood educators. International Journal of Educational Research, 115, 102042. https://doi.org/10.1016/j.ijer.2022.102042
  18. Sims, S., Fletcher-Wood, H., O’Mara-Eves, A., Cottingham, S., Stansfield, C., Goodrich, J., Van Herwegen, J., & Anders, J. (2025). Effective teacher professional development: New theory and a meta-analytic test. Review of Educational Research, 95(2), 213–254. https://doi.org/10.3102/00346543231217480
  19. 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
  20. UNESCO. (2021). Recommendation on the ethics of artificial intelligence. UNESCO. https://unesdoc.unesco.org/ark:/48223/pf0000381137
  21. 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
  22. Yang, J., Zhang, Y., Zhou, Y., & Ji, Y. (2023). What are the core features of professional learning community in Chinese preschool teachers’ perspectives: Based on grounded theory analysis. Frontiers in Psychology, 14, 1177321. https://doi.org/10.3389/fpsyg.2023.1177321
  23. Yang, W., Huang, R., Su, Y., Zhu, J., Hsieh, W.-Y., & Li, H. (2022). Coaching early childhood teachers: A systematic review of its effects on teacher instruction and child development. Review of Education, 10(1), e3343. https://doi.org/10.1002/rev3.3343
  24. Zhang, H., & Wang, Q. (2025). How could GenAI work on in-service teachers’ knowledge building process? An empirical study based on epistemic network analysis. International Journal of Educational Technology in Higher Education, 22, Article 47. https://doi.org/10.1186/s41239-025-00544-y