Are AI-powered learning platforms really more engaging?

A group of 23 students tested three popular online learning platforms – Khan Academy, Coursera, and Codecademy – to see how AI features affect user experience. The experiment focused on learning HTML and measured task completion time, satisfaction, engagement, and answer accuracy. Codecademy stood out for speed: students finished exercises fastest. Khan Academy, however, earned the highest satisfaction scores thanks to a friendlier interface.

Surprisingly, the AI-driven adaptive features were almost invisible. Participants felt that the platform's core interactivity (instant feedback, hands‑on exercises) mattered more than personalization algorithms. This led to a key conclusion: there is a trade‑off between learning efficiency and engagement, mediated by the discoverability and perceived value of AI features.

Why aren’t platforms fully leveraging AI? The authors recommend making adaptive functions more visible, intuitive, and embedded in the learning flow. Only then can students enjoy a more personalized learning path without losing motivation.

The study employs a mixed-methods design, merging quantitative metrics (completion time, accuracy) with qualitative insights (semi-structured interviews). Selecting HTML as the test content is strategic: it is a markup language with simple syntax yet requires conceptual understanding of tags, attributes, and hierarchical structure. This isolates interface effects without introducing overly complex cognitive variables.

The examined platforms implement distinct adaptive models. Khan Academy uses a Bayesian Knowledge Tracing algorithm that estimates the probability a learner has mastered a concept based on prior responses. Coursera relies on a hybrid recommendation system, blending collaborative filtering with content analysis of the course. Codecademy adopts a more reactive approach, delivering immediate feedback and contextual hints based on predefined rules.

Findings reveal that the mere presence of algorithms does not guarantee perceived improvement. "Discoverability"—the user’s ability to recognize and understand adaptive features—emerges as critical. When AI functions are hidden or poorly explained, students default to traditional interface cues, diminishing the impact of adaptivity. This aligns with cognitive load theory, which posits that unfamiliar elements can increase extraneous load, hindering learning.

To optimize platform efficacy, the authors propose three guidelines: (1) make AI features explicit through short tutorials and progress visualizations; (2) embed adaptive suggestions directly into the workflow, avoiding interruptions; (3) tailor transparency levels to user profiles, offering an "expert mode" for those seeking finer control. Implementing these strategies could narrow the efficiency‑engagement gap, fully leveraging artificial intelligence’s potential in digital education.