Last updated: July 2026 | By LearnPath Team
What Is LearnPath?
LearnPath is an AI-powered learning platform that transforms free YouTube content into structured, personalized learning journeys. You tell the system what you want to learn, and it builds a complete course for you - curated videos, exercises generated from actual video transcripts, adaptive branching that responds to your performance, and spaced repetition that ensures long-term retention. Building a path takes about 30 to 90 seconds.
YouTube already hosts millions of hours of high-quality educational content from expert creators. The problem has never been a lack of material. The problem is that YouTube was not designed to be a learning management system. There is no curriculum, no assessment, no progress tracking, and no way to adapt the experience based on how well you understand the material. LearnPath adds all of that on top of the content that already exists.
This post walks through every step of how LearnPath works, from the moment you type in a topic to the moment you complete a fully adaptive learning path.
Step 1: Onboarding - Tell LearnPath What You Want to Learn
The process begins with a short onboarding wizard that collects five pieces of information about your learning goal.
Topic. What do you want to learn? This can be as broad as "machine learning" or as specific as "React Server Components." LearnPath works with any topic that has YouTube coverage.
Current level. Are you a complete beginner, do you have some experience, or are you already advanced? This determines the starting difficulty of the content the AI selects.
Learning goals. What do you want to be able to do when you finish? Build a project? Pass an exam? Understand theory? Your goals shape whether the AI prioritizes hands-on tutorials, conceptual explanations, or exam-prep material.
Preferences. Do you prefer short, focused videos or long deep-dives? Visual diagrams, code-alongs, or lecture-style explanations? These preferences influence how the AI scores candidate videos.
Preferred creators. If you already follow specific YouTube channels, you can list them. The AI will prioritize content from these creators while still supplementing with other high-quality sources.
This onboarding data creates a learner profile that the AI references for every decision throughout your path - from initial video selection to branching logic to exercise difficulty.
Step 2: AI Video Curation - Finding the Right Content
Once onboarding is complete, LearnPath's AI runs a multi-pass content curation pipeline. This is not a simple YouTube search.
YouTube Search and Transcript Retrieval
The system queries the YouTube Data API with search terms derived from your topic, level, and goals, retrieving video metadata including titles, descriptions, view counts, channel information, and durations. For each candidate, LearnPath retrieves the full transcript - manually uploaded captions first (more accurate), then auto-generated captions, then the video description as a last resort.
The transcript is the most important input. Unlike titles and descriptions, which are often written for SEO, the transcript reveals exactly what a video teaches, how it is structured, and at what level.
Multi-Pass Gemini Filtering
LearnPath sends video metadata and truncated transcripts to Google's Gemini AI for evaluation. The AI scores each video on content accuracy, teaching clarity, relevance to the learner's goals, prerequisite alignment, and production quality. Videos below the quality threshold are filtered out. The remaining videos are ranked by composite score, with preference given to the learner's preferred creators.
The result is a curated set of high-quality, level-appropriate videos that form the raw material for the learning path.
Step 3: Curriculum Generation - Module Structure and Skill Mapping
With ranked videos in hand, the AI constructs a structured curriculum - not a flat playlist, but a branching tree where each node represents a video plus its associated exercises.
The AI analyzes transcripts and maps the specific skills and concepts each video covers. This skill map determines sequencing: foundational concepts come before advanced ones, and no critical prerequisite is skipped. If you are learning Python, the system ensures you encounter for loops before list comprehensions.
The tree also includes deduplication logic. If two videos cover substantially the same material, the AI will not include both in the same branch. If you have already demonstrated mastery of a concept, the system will not route you to redundant content.
Step 4: Six Exercise Types - Testing Real Understanding
After you watch each video, LearnPath generates exercises directly from the video's transcript. These are not generic questions from a question bank. They test your understanding of the specific material you just watched. LearnPath offers six distinct exercise types. Multiple-choice quizzes are available on every plan; the other five types unlock on Pro. You can explore all plan details on the pricing page.
Multiple-Choice Quizzes (available on Free and Pro)
The AI generates multiple-choice questions at varying difficulty levels, designed to test genuine comprehension rather than surface-level keyword recall. Each question includes plausible distractors based on common misconceptions. Quizzes are available to every user and serve as the primary assessment driving adaptive branching.
Fill-in-the-Blank and Flashcard Drills (Pro)
Fill-in-the-Blank exercises present statements from the video with key terms removed. You recall and type the correct term from memory. The system accepts alternative phrasings and common synonyms. This targets active recall - a more demanding cognitive task than recognition.
Flashcard Drills present term-definition pairs extracted from the video. You see one side, attempt to recall the other, and rate your confidence as "got it," "almost," or "missed." This self-assessment feeds into the spaced repetition system.
Concept Matching, Sequencing, and True/False with Explanations (Pro)
Concept Matching presents two columns - concepts and definitions - and asks you to draw correct connections. This tests associative understanding, a deeper form of comprehension than isolated recall.
Sequencing Challenges present steps or concepts in scrambled order for you to arrange correctly. This is particularly effective for procedural knowledge: algorithms, workflows, or any topic where order matters.
True/False with AI Explanations presents statements for you to evaluate, then provides detailed AI-generated explanations citing specific parts of the video. This combines assessment with teaching in the same interaction.
All six types are generated from the actual transcript, ensuring every question is directly relevant. See the features page for a full breakdown of what each plan includes.
Step 5: Adaptive Branching - Your Path Responds to Your Performance
After you complete a quiz for a node, the system evaluates your score against a pass threshold and makes a branching decision that determines what you learn next. This is what makes LearnPath fundamentally different from a static course.
Pass. You cleared the bar and demonstrated solid understanding. The system moves you forward to the next logical step in the skill progression, so no time is wasted on material you already know.
Reinforce. You fell short of the pass mark. Rather than pushing you into content that builds on concepts you have not grasped, the system inserts supplementary material before you move on - an easier video, or extra practice on the same concept explained from a different angle. This is not a penalty. It is the system ensuring you build a solid foundation first.
Because the path branches on your actual results, two learners who start the same topic can end up taking very different journeys through the material, each optimized for their individual understanding. The AI considers your performance across the path, not just an isolated question.
Step 6: Spaced Repetition - Retaining What You Learn
Learning something once is not enough. Without periodic review, newly acquired knowledge decays rapidly. LearnPath addresses this with a spaced repetition system based on the SM-2 algorithm.
When you complete exercises, questions you answered incorrectly are automatically converted into review cards. The SM-2 algorithm tracks how easily you recall each card and schedules the next review at the optimal interval. Cards you find easy get pushed further out (days, then weeks, then months). Cards you struggle with come back sooner.
Your dashboard shows how many review cards are due. Review sessions present cards from across your entire learning history, weighted toward concepts you found difficult. Over time, spaced repetition produces two to three times better long-term retention compared to one-time study.
The exercises you complete during your path are not throwaway assessments. They become a growing library of review material that keeps your knowledge fresh indefinitely.
The Two Plans
LearnPath has two plans, and you can start for free with no credit card.
Free - 1 learning path with the entire first step unlocked (typically 3 to 5 curated video lessons, each with its own transcript-based quiz), multiple-choice quizzes, daily review cards, and progress tracking with XP and streaks. It is a real learning experience on a real skill, meant to show you whether LearnPath fits how you learn. Finishing the first step shows the upgrade wall.
Pro ($12.99/month, or $8.99/month billed annually) - Unlimited learning paths, all six exercise types, unlimited review cards, the full analytics dashboard, completion certificates you can share on LinkedIn, and the ability to add your own YouTube videos to a path.
Compare all features side-by-side on the comparison page, or see current pricing on the pricing page.
Explore Before You Build a Path
You do not have to start from a blank box. Two features help you find the right content first:
- Best Video Finder at /find takes a single question or skill and returns the single best YouTube video to watch for it, with the AI's reasoning - useful when you want one great video rather than a whole path.
- Discover pages at /discover are public, pre-built topic pages that show what a LearnPath course on a subject looks like before you sign up.
Frequently Asked Questions
How long does it take to generate a learning path?
Path generation typically takes 30 to 90 seconds. The system searches YouTube, retrieves transcripts, runs the AI filtering pipeline, and constructs the tree structure. Exercises are generated on demand as you reach each node, so there is no additional wait before your first video.
Does LearnPath work for any subject?
LearnPath works with any subject that has meaningful YouTube coverage. Users have created paths for programming languages, music theory, history, cooking techniques, foreign languages, mathematics, and more. The AI adapts its exercise generation to the domain - it creates different question types for a cooking video than for a calculus lecture.
Can I influence which videos are selected?
Yes. During onboarding, you specify preferred YouTube creators whose content is prioritized. The branching system also responds to your behavior - strong performance on certain content types is factored into future selections.
How is LearnPath different from a YouTube playlist?
A playlist is static: same videos, same order, no assessment, no adaptation. LearnPath adds AI-curated video selection matched to your level, six exercise types generated from transcripts, adaptive branching based on performance, spaced repetition for long-term retention, and progress tracking with XP, streaks, and completion certificates. The result is closer to a private tutor than a playlist.
What happens if I score poorly on exercises?
The system branches you toward supplementary material that covers the same concepts from a different angle. You are never stuck or penalized. The adaptive branching ensures you build understanding at your own pace before advancing to more complex topics.