Teaching Prompt-Based Programming with LLMs: A 45-Minute Lesson with Guided Practice for End-User Programmers

Synthesis: This study by Tran, Marwan & Price (2026) introduces and evaluates a 45-minute structured lesson on prompt-based programming, a new modality enabled by LLMs where users express computational goals through natural language. The lesson design incorporates guided practice principles and targets end-user programmers with limited formal training. Results show significant pre-to-post gains in prompt quality, code correctness, and Self-Efficacy, supporting the case that AI Literacy interventions need not be lengthy to be effective. The work connects to broader conversations about how Large Language Models (LLMs) tools change the skills required for programming — shifting emphasis from syntax to prompt engineering. By focusing on end-user programmers rather than CS students, the study expands the scope of STEM Education research to include lifelong and professional learning contexts. It also raises questions about the Teaching in an AI-mediated classroom, where instructors must now teach prompt design alongside or instead of traditional coding concepts. Future work could explore how such lessons scale across Higher Education institutions and whether the gains translate to real-world programming tasks.

What this means for practice

• Instructors. Budget a single class period for prompting rather than a whole unit, and expect modest rather than transformative gains: the experimental group improved +10.8 percentage points on the post-test against +1.1 for the control, an estimated 0.87 additional problems solved out of 8.
• Instructors. Teach a structured template that makes students state the input and its type, the output, and the transformation, instead of leaving prompting to trial and error.
• Instructors. Practice prompts in the same format students will be assessed in, since part of the experimental advantage may reflect familiarity with the assessment format rather than prompting skill alone.
• Instructors. Keep code comprehension and tracing work in the course alongside prompting: the control lesson's explain-in-plain-English exercises may have helped students interpret LLM-generated code and thereby narrowed the gap.
• Researchers. Examine the prompts students actually write, not only their scores, because the study did not analyze prompt content and cannot explain which strategies produced the gains.

Limitations

• The intervention lasts 45 minutes — a single class period — and the comparison is a business-as-usual CS lab activity (code tracing) rather than a no-instruction condition.
• All sessions were led by the first author, who also designed the curriculum, introducing potential experimenter expectancy bias even with structured slides and facilitator notes.
• The performance effect was not statistically significant (B = 0.756, 95% CI [−0.020, 1.533], p = 0.056) and less than one additional problem solved on average, with the model fitted to 106 observations across 53 participants.
• Sessions ran in both in-person and online formats (experimental: 2 in-person, 5 online; control: 2 in-person, 6 online), and only short-term pre-to-post gains were measured, so persistence and transfer of prompting skill remain untested.

Citation

Keith Tran, Samiha Marwan, Thomas Price (2026). Teaching Prompt-Based Programming with LLMs: A 45-Minute Lesson with Guided Practice for End-User Programmers.