Conceptualizing pre-service teachers' readiness for AI integration into teaching practices: An intelligent-TPACK approach

Synthesis: Reyes-Rojas, Díaz, Ruz-Reveco, Castro, and Reyes-González (2026) adapt the intelligent-Technological Pedagogical Content Knowledge (TPACK) (iTPACK) instrument originally developed by Celik (2023) for in-service teachers to measure Professional Development' readiness to integrate AI into their training context. Conducting a quantitative cross-sectional survey of 366 pre-service teachers drawn from eleven universities in Chile, they translated the 27-item instrument into Spanish through a two-translator cross-cultural adaptation and analyzed responses with Confirmatory Factor Analysis (CFA) using both Maximum Likelihood and Weighted Least Squares Mean and Variance Adjusted (WLSMV) estimators, which are better suited to ordinal Likert data. The adapted model achieved excellent fit after removing three items (CFI = 0.997, TLI = 0.997, RMSEA = 0.028, SRMR = 0.038), and non-parametric Quade tests controlling for age and academic level revealed statistically significant differences in iTPACK dimensions by academic progress level and by type of institution (public vs. private), while gender, geographic location, and type of major showed no noteworthy differences. The study extends the iTPACK conceptualization to the pre-service population, providing a validated Spanish instrument that foregrounds the framework's Ethics domain and surfacing key areas for future curriculum development and support in Professional Development.

Key Findings

Instrument validation via CFA. The adapted Spanish iTPACK instrument retained the five dimensions of the original (iTK, iTPK, iTCK, iTPACK, and ETHICS). A Confirmatory Factor Analysis using both MLE and WLSMV estimators — the latter preferred because it handles the ordinal nature of Likert responses — produced a strong fit after dropping three items (original items 6 and 7 from the iTPK factor and item 26 from the ETHICS factor, keeping every factor above the three-item minimum): robust CFI = 0.997, TLI = 0.997, RMSEA = 0.028 (90% CI 0.024–0.033), and SRMR = 0.038. Standardized factor loadings ranged from 0.760 to 0.914 and were all significant at p < 0.001. All analyses ran in JASP 0.19.3, with Quade tests implemented in Python (pandas, scipy.stats, statsmodels).

Pre-service teachers as overlooked stakeholders. The authors argue that pre-service teachers are frequently excluded from the development of AI-integration tools and instruments, even though they will be crucial agents in integrating AI-based tools into the new educational landscape. A key literature gap is the absence of instruments that (1) measure iTPACK, (2) include ethical factors, (3) are validated for pre-service teachers, and (4) are validated in Spanish — all four gaps this study addresses.

Demographic relationships via Quade tests. After Shapiro–Wilk tests indicated non-normal distributions for all factors (p < 0.001), non-parametric ANCOVA (Quade test) controlling for age and academic progress was used to compare groups. Statistically significant effects emerged for academic progress level and type of institution: public-university students (n = 251) scored higher than private-university students (n = 115) across all five dimensions (e.g., ETHICS F = 35.38, p < 0.001; iTK F = 13.87, p < 0.001), and upperclassmen reported higher readiness than underclassmen (e.g., ETHICS 4.77 vs. 4.12, p = 0.001). Gender, geographic location (capital vs. region), and type of major (STEM vs. non-STEM) did not demonstrate noteworthy differences. The sample was predominantly 23–30 years old (47.27%), female (65.03%), senior students (61.20%), and enrolled in public institutions (68.58%).

Ethical dimension and factor structure. The ETHICS factor showed the highest correlation with iTPACK (ρ = 0.833 for iTCK–iTPACK; ETHICS correlated most strongly with iTPACK and least with iTK), suggesting that responsible, accountable decision-making around AI tools may be enhanced when such tools are integrated into teaching and learning rather than approached solely through their technological properties. This mirrors Celik's original iTPACK findings and underscores the Ethics and AI Literacy dimensions central to the framework.

Implication. Validating readiness instruments for the pre-service population is a necessary step toward designing Professional Development curricula that prepare future teachers to integrate AI responsibly — a central concern of the knowledge base's teacher-education concept. The authors also recommend future work with SEM and additional control variables, and note the study was approved by the Ethics Committee of the Pontificia Universidad Católica de Chile (ID 250,310,001), with the instrument itself available in Appendix A.

What this means for practice

• Faculty developers. Use the validated Spanish instrument (27 items, 3 dropped, five domains retained) to diagnose pre-service readiness before designing AI training, and begin with the ETHICS domain, which correlated most strongly with integrated iTPACK (ρ = 0.833).
• Faculty developers. Front-load AI integration for underclassmen: upperclassmen reported higher readiness than underclassmen (ETHICS 4.77 vs. 4.12, p = 0.001).
• Instructors. Teach AI through pedagogical and ethical integration rather than technological properties, since the ETHICS factor tracked iTPACK more closely than technical knowledge.
• Administrators. Treat the public–private gap as an equity problem: public-university students scored higher on all five dimensions (e.g., ETHICS F = 35.38; iTK F = 13.87), so private institutions need targeted support.

Limitations

• Cross-sectional, voluntary online survey (Google Forms) of 366 pre-service teachers at 11 Chilean universities: it measures perceived knowledge, not observed classroom integration, and cannot establish causal direction.
• The instrument is a Spanish adaptation of an in-service measure, and three original items were removed after initial fit problems (items 6 and 7 from iTPK, item 26 from ETHICS), so the reported model reflects a post hoc modification.
• Data come from one national system with one translation, and the sample skews female (65.03%) and senior (61.20%), so the five-factor structure is untested in other languages and teacher-education systems.
• Group comparisons rest on collapsed categories (underclassmen vs. upperclassmen; public vs. private) analyzed with non-parametric Quade tests, since Shapiro–Wilk tests showed non-normal distributions for every factor (p < 0.001).

Citation

Reyes-Rojas, J., Díaz, B., Ruz-Reveco, C., Castro, A., & Reyes-González, D. (2026). Conceptualizing pre-service teachers' readiness for AI integration into teaching practices: An intelligent-TPACK approach. Computers and Education Open, 100320. https://doi.org/10.1016/j.caeo.2025.100320