Chemistry teachers' research competence following AI-supported professional development: a mixed-methods pre-post study

Synthesis: Teachers are increasingly expected to act as practitioner-researchers, which makes Teacher AI Competency part of the research dimension of professional expertise. This mixed-methods study traced 105 chemistry teachers through a six-week, AI-supported professional development program built around generative AI tools. Research competence correlated positively with social-emotional monitoring practices (r = .664, p < .01) and readiness to use AI (r = .622, p < .01), and all measured dimensions improved significantly from pre-test to post-test (p < .001), with Cohen's dz values ranging from 0.47 to 0.63. Qualitatively, teachers shifted from seeing AI as a tool for information retrieval and writing to viewing it as support for research planning and data-informed decision-making in AI-assisted educational research. The study proposes an empirically informed conceptual model integrating teacher inquiry, social-emotional monitoring, and AI-supported research practices for chemistry education, grounding AI-supported research practices in teachers' everyday inquiry and positioning AI as an analytic partner for evidence-informed Pedagogies and Teaching Strategies.

Key Findings

1. Data came from 105 chemistry teachers, using closed-ended questionnaire scales and two open-ended questions administered before and after a six-week AI-supported professional development program.
2. Research competence correlated positively with social-emotional monitoring practices (r = .664, p < .01) and with readiness to use AI (r = .622, p < .01).
3. Statistically significant pre-post improvements appeared across all measured dimensions (p < .001), with Cohen's dz values ranging from 0.47 to 0.63.
4. The largest standardized change was in teachers' research competence (dz = 0.63), whose mean rose from 3.50 (SD = 0.88) to 4.16 (SD = 0.52).
5. Qualitatively, teachers' perception of AI shifted from a tool for information retrieval and writing to one supporting research planning and data-informed decision-making.
6. The study proposes an empirically informed conceptual model integrating teacher inquiry, social-emotional monitoring, and AI-supported research practices.

Research competence and AI in teacher expertise

Research competence has become a core dimension of teachers' expertise, covering the ability to identify educational problems, formulate research questions, and interpret empirical evidence. The demands are acute in chemistry, a subject that integrates abstract theory, macroscopic observations, submicroscopic explanations, and symbolic representations and frequently produces persistent learning difficulties. In this framing, teachers act as reflective practitioner-researchers rather than transmitters of disciplinary knowledge. The authors argue that AI can support nearly every stage of the research process, from identifying and synthesizing literature to analyzing data and interpreting findings; this study supplies the missing empirical account of how AI-supported professional development changes chemistry teachers' research competence.

A six-week program for classroom-based inquiry

The program ran in stages over six weeks, one per research stage, each combining three parts: review of theoretical material with demonstrations of relevant AI tools, authentic research activity inside teachers' own schools and chemistry classrooms, and collaborative reflection with professional dialogue. Teachers worked from a self-identified pedagogical problem through literature review, instrument design, data collection, statistical analysis, and a mini research report, with tools such as ChatGPT, Perplexity, Scite, and Julius AI. AI functioned as a cognitive and methodological scaffold rather than a substitute for professional expertise. The evaluation used a single-group pre-test/post-test design with an embedded qualitative component — closed-ended scales plus two open-ended questions — analyzed with Pearson correlations, paired-samples t-tests, and qualitative content analysis under a mixed-methods logic.

Results: correlations and pre-post gains

Correlation analysis revealed positive relationships among all five questionnaire constructs (p < 0.01), with coefficients ranging from 0.443 to 0.671. Research competence correlated most strongly with social-emotional monitoring practices (r = .664, p < .01) and with readiness to use AI (r = .622, p < .01). After the program, mean scores rose on every scale and paired-samples t-tests showed statistically significant improvements on all dimensions (p < .001), with Cohen's dz values ranging from 0.47 to 0.63 — approximately medium effects. The largest change was in research competence (dz = 0.63; from 3.50, SD = 0.88, to 4.16, SD = 0.52), followed by readiness to use AI (dz = 0.54), and post-test standard deviations were consistently lower, indicating greater consistency after the intervention.

A conceptual model for developing research competence

The qualitative analysis documented a shift in how teachers characterized AI: before the intervention they emphasized preparing lessons, developing materials, and searching for information; afterward they increasingly described AI as a methodological partner supporting research planning, data collection, analysis, and reporting. The share of responses applying AI across teaching and research activities rose from 12 (11.4%) before the program to 38 (36.9%) after it. Drawing on CASEL's social and emotional competence framework, the study proposes an empirically informed conceptual model integrating teacher inquiry, social-emotional monitoring, and AI-supported research practices, with AI as analytic support for evidence-informed pedagogical decisions rather than a replacement for judgment.

What this means for practice

• Instructors. Use AI as a cognitive and methodological support in teachers' research, not a substitute for expertise; the gains were associated with a program that also included research-methods training, classroom inquiry, and collaborative reflection.
• Faculty developers. Structure professional development around authentic classroom inquiry — one research stage per week, with AI tools matched to each stage from literature review through data analysis to a mini research report.
• Administrators. Expect the largest shifts in teachers' research competence and readiness to use AI (dz = 0.63 and 0.54), but treat single-group pre-post results as changes associated with participation rather than causal proof.

Limitations

• The sample comprised 105 in-service chemistry teachers from the South Kazakhstan region, ages 24 to 59, recruited by purposive sampling, limiting generalization to other subjects, contexts, or national systems.
• The single-group pre-test/post-test design had no comparison or control group.
• No direct student-level social-emotional measures were collected and no AI models analyzed students' social-emotional states; conclusions rest on teachers' perceptions and reported practices.
• The multicomponent program prevents isolating AI's independent contribution from research-methods instruction, classroom inquiry, and collaborative reflection.
• Two participants did not respond to Question 35 after the intervention, so post-program percentages were calculated from 103 valid responses.

Citation

Ussenbay, A., Abdurazova, P., Raiymbekov, Y., Kerimbayeva, K., Ualikhanova, B., & Usmantayev, A. (2026). Chemistry teachers' research competence following AI-supported professional development: a mixed-methods pre-post study. Frontiers in Education, 11:1961742.