How Do AI Tutoring Platforms Decide What to Teach a Student Next?

A Map of Skills, Plus Real-Time Performance Data

AI tutoring platforms generally don’t invent a learning path from scratch for each student. Instead, they’re built on top of a predefined structure — often called a knowledge graph, skill tree, or curriculum map — that lays out the individual skills within a subject and how they depend on each other. Long division, for example, might be mapped as depending on multiplication fluency, which in turn depends on basic addition. This structure is typically designed by curriculum experts and educators before any student ever uses the platform.

Once a student starts working, the platform layers their real-time performance on top of that map. Correct and incorrect answers, how long a student takes to respond, and the specific type of errors they make all feed into an ongoing estimate of where that student currently sits on the skill map — which skills they’ve likely mastered, which are in progress, and which they haven’t been introduced to yet.

How the “Next” Decision Actually Gets Made

When it’s time to select the next thing to show a student, most systems combine two considerations: the prerequisite structure of the skill map, and a mastery estimate for the student’s current skill. If a student demonstrates consistent success on a skill — not just a single lucky correct answer, but a pattern across several questions — the system typically moves them toward the next skill in the sequence for which the prerequisites have been met. If a student struggles, the system usually responds in one of two ways: offering additional, sometimes easier, practice on the current skill, or stepping back to review an earlier prerequisite skill that may be the actual source of the difficulty.

This is a meaningfully different approach from a fixed curriculum where every student advances through the same material on the same school-year schedule regardless of individual mastery.

Why the Underlying Map Still Matters a Lot

Because the quality of these recommendations depends heavily on how well the underlying skill map and mastery thresholds were designed, different platforms can produce noticeably different learning paths even for students with similar performance. A platform with a finely grained skill map and conservative mastery thresholds might move a student more slowly and thoroughly than one with a coarser map and looser thresholds. This is a meaningful, if often invisible, design decision that shapes the actual educational experience a student gets.

Bottom Line

AI tutoring platforms decide what to teach next by combining a predefined map of skills and prerequisites with a student’s ongoing performance data, moving students forward when they show consistent mastery and looping back to review or reinforce when they struggle — with the quality of that experience depending significantly on how carefully the underlying skill map was designed.

Go deeper

• What is a micro credential and how does it differ from a full certification?
• What is the difference between a specialization and a single course on major learning platforms?
• How Do You Use AI to Learn a New Skill Effectively?