Brett LindenbergAdaptive learning examples include Khan Academy's mastery system, ALEKS in college math, and Duolingo's lesson engine. Each one changes the next question, lesson, or review based on what the learner just did. A missed answer brings targeted review, and a run of correct answers brings harder work or a chance to skip ahead. An adaptive learning platform is the software that makes that choice for each learner, using rules an author writes or a model that estimates what the learner knows. In the short video below, your first answer picks the lesson that plays next.
What adaptive learning is and how platforms adapt
Adaptive learning is instruction that adjusts to evidence from each learner. The evidence is usually answers, and some platforms also use response time or a confidence rating. The adjustment can be the next question, its difficulty, or a review before the learner moves on. A self-paced online course where everyone sees the same lessons in the same order does not adapt, however polished it is.
Platforms decide what comes next in one of three ways, and many products combine them.
| Type | How it decides | Examples | Watch for |
|---|---|---|---|
| Rule-based paths | An author writes if-then rules, such as "a score under 70% opens the review lesson" | Canvas MasteryPaths, Moodle access restrictions, Mindstamp video branching | Rules only cover the mistakes you planned for |
| Learner model | Software estimates what each learner knows and picks the item most likely to help | ALEKS, Duolingo, MAP Growth | Needs a large bank of tagged questions, and the logic is hard to inspect |
| Generative AI tutor | A language model answers the learner's questions and adjusts its explanations in conversation | Khanmigo, Coursera Coach | May not track mastery or change the course order |
"AI adaptive learning" usually means the second or third type. Duolingo's Birdbrain model, introduced in 2020, predicts whether a learner will get a given exercise right, and the lesson builder uses that prediction to pick exercises at the right difficulty (Duolingo, 2020). Generative tutors work differently. Khanmigo, Khan Academy's AI tutor, guides students toward an answer with questions instead of giving it to them. Coursera adds an assistant called Coursera Coach to its courses. Coach answers questions about the material and writes practice questions, and Coursera reported in December 2024 that learners who used it completed 11.6% more course items per hour (Coursera). Tutors like these change the explanation. The course sequence stays the same.
Adaptive learning platforms compared
The table lists 14 adaptive learning examples by setting. We checked each vendor's own pages in October 2026. The sections after it explain how each one adapts.
| Platform | Setting | Adapts to | What changes for the learner |
|---|---|---|---|
| Khan Academy | K-12 and early college, free | Practice, quiz, and unit test results | Skill mastery levels go up or down |
| DreamBox Math | Grades K to 8 math | Answers and how the student solves each problem | The next lesson and its difficulty |
| IXL Real-Time Diagnostic | K-12 math and English language arts | Diagnostic answers and later practice | An action plan of skills to practice next |
| MAP Growth | K-12 assessment | Each answer during the test | The next question gets harder or easier |
| CENTURY | Ages 6 to 16, English, maths, and science | Self-marking question results | Recommended next lessons |
| ALEKS | K-12 and college math, chemistry, statistics, accounting | Initial and periodic Knowledge Checks | Only topics the learner is ready to learn |
| Pearson MyLab and Mastering | College courses | Answers plus a confidence rating | Questions cycle until each one is mastered |
| Knewton Alta | College STEM, economics, psychology | Assignment answers | Review of missing prerequisite skills |
| SmartBook | College reading assignments | Answers to reading questions | Mastered concepts drop out, weak ones repeat |
| Duolingo | Language learning | Each exercise answer | Exercise difficulty within a lesson |
| Santa | TOEIC and TOEFL prep | Diagnostic and practice answers | A predicted score and targeted practice |
| Area9 Rhapsode | Workplace, higher education, K-12 | Answers plus a confidence rating | What to study next, including gaps the learner did not suspect |
| Skillwell Adapt | Workplace and higher education | Prior-knowledge checks and performance | Learners test out of concepts they know |
| Mindstamp (our product) | Training and education video | Answers, clicks, and variables in the video | The next scene, a review section, or a follow-up question |
Adaptive learning examples in K-12
Khan Academy
Khan Academy is a nonprofit with 840+ free courses and 227 million registered users, according to its about page. Its adaptivity comes from mastery learning. Each skill sits at one of four levels, from Attempted to Mastered, and the level moves up or down as the student answers practice questions, quizzes, and unit tests. A skill a student mastered in October can drop if the student misses it on a December unit test. Teachers see these levels for every student. Khanmigo is free for teachers, while student access needs a paid plan or a district partnership.
DreamBox Math
DreamBox Math, now sold by Discovery Education, supplements core math instruction for grades K to 8 in class or at home. Discovery Education says the program adapts in real time to a student's answers and to how the student approaches and solves each problem. Lessons are available in English and Spanish.
IXL Real-Time Diagnostic
The IXL diagnostic places a student in math and English language arts in about 45 minutes per subject. It then builds a plan of the skills the student is ready to learn next. Levels and recommendations keep updating as the student practices on IXL.
MAP Growth
MAP Growth from NWEA is an adaptive test for kindergarten through grade 12 in math, reading, language arts, and science. A correct answer brings a harder question and a wrong answer an easier one. Scores land on NWEA's RIT scale, which does not depend on grade level, so a school can track one student's growth across years. NWEA says 13 million students in 35,900 schools take it.
CENTURY
CENTURY is a UK platform for students aged 6 to 16. It pairs thousands of teaching videos and self-marking questions with a recommended path for each student, mainly in English, maths, and science. CENTURY says schools in more than 70 countries use it.
If you are choosing adaptive learning software for a private or independent school, decide two things before you compare vendors. First, pick the subject where skill levels in one class vary the most. Second, decide whether you need a diagnostic test that reports levels (IXL, MAP Growth) or daily practice that adapts (DreamBox Math, Khan Academy). Then ask how a teacher will act on the report the next morning.
What options exist for adaptive learning platforms in higher education?
Four common options for adaptive learning in higher education come from textbook publishers. You adopt the courseware along with the course materials, so the choice often follows the textbook decision.
ALEKS
ALEKS, which McGraw Hill acquired in 2013, uses knowledge space theory. A short adaptive Initial Knowledge Check places each student among what ALEKS calls trillions of possible knowledge states. ALEKS then offers only the topics the student is ready to learn, and periodic Knowledge Checks confirm what the student retained. McGraw Hill says that when ALEKS judges a student ready for a topic, the student masters it at least 90% of the time. Course products cover math, chemistry, statistics, and accounting.
Pearson MyLab and Mastering
Pearson's MyLab and Mastering homework platforms include Dynamic Study Modules. Students answer a set of questions and mark each answer "I'm sure," "I think so," or "I don't know." The module keeps cycling through the questions until the student shows mastery of each one (Pearson, 2024). The confidence rating separates a lucky guess from real knowledge, which a plain score cannot do.
Knewton Alta
Wiley bought Knewton's assets in 2019 and still sells Knewton Alta as adaptive courseware for STEM. The catalog also covers economics and psychology. Many of its math courses include corequisite support, which reviews prerequisite skills while students work on college-level content.
SmartBook in McGraw Hill Connect
SmartBook turns a reading assignment into short cycles of questions on no more than five concepts at a time. Once a student shows they understand a concept, its questions stop appearing. If the student keeps struggling, related questions come back. Instructors choose the topics, the depth of coverage, and how long students spend.
A fifth option is already on most campuses: the rule-based paths built into the LMS, covered further down.
Adaptive learning in language apps and test prep
Duolingo
Duolingo had 140.6 million monthly active users and 58.7 million daily active users in the quarter that ended June 30, 2026, according to its shareholder letter. Its Birdbrain model estimates how likely each learner is to answer each exercise correctly, and the app uses those estimates to choose exercises that are hard enough to teach without being discouraging.
Santa
Santa is a TOEIC and TOEFL prep app from the Korean company Riiid, which renamed itself Socra AI in September 2025. The app predicts a learner's test score from their answers and focuses practice on the learner's goal.
Adaptive learning in workplace training
Workplace programs adapt for a practical reason: an experienced employee should not repeat material they already know, and a new hire should not skip it. The common pattern is a check before each topic that lets people test out.
Area9 Rhapsode
Area9 Lyceum's Rhapsode platform asks learners to rate their confidence before the answer is revealed. A confident wrong answer exposes a gap the learner did not know about, which matters in safety and clinical training. Area9 sells to employers, universities, and schools, and it claims a 50% cut in time for first-time learning.
Skillwell Adapt
The former Realizeit adaptive platform now sits under the Skillwell name as Skillwell Adapt. Its Determine Knowledge feature checks what a learner already knows so they can test out of concepts and spend their time on gaps. Skillwell says this saves learners 40% of their time. It serves corporate training and higher education.
Mindstamp interactive video
Mindstamp, our product, adds rule-based paths to video you already have. You place a question or a clickable choice where a concept is taught. Each answer can jump to another part of the video, switch to another video, or show a short explanation. The demo at the top of this page uses three hotspots, three routes, and a follow-up question on each route. Reports show each learner's answers and path. Mindstamp does not build a statistical model of each learner. You write the routes, so you can read, test, and change every one of them. Our adaptive learning page shows the setup for a lecture or lesson.
Design patterns you can copy from these platforms
You do not need any of these products to use their best ideas. Each pattern below fits a quiz in your LMS as well as an interactive video.
| Pattern | Seen in | How to apply it |
|---|---|---|
| Check before you teach | ALEKS, IXL, Skillwell Adapt | Open with a short check and let a correct answer skip the matching section |
| Ask for confidence | Pearson Dynamic Study Modules, Area9 Rhapsode | Add "How sure are you?" after key questions and review confident wrong answers first |
| Retire what is mastered | SmartBook | Stop asking about a concept once the learner gets it right |
| Recheck later | ALEKS Knowledge Checks, Khan Academy unit tests | Retest early skills at the end, and let a level go down after a miss |
| Step difficulty up or down | MAP Growth, Duolingo | Follow a correct answer with a harder case and a miss with an easier one |
Can an LMS do adaptive learning?
The three LMSs below support rule-based adaptive paths. None of these features builds a learner model on its own, but score-based release covers many real needs.
- Canvas MasteryPaths releases different assignments or pages based on a student's score on a source assignment. The default score ranges start at 70%, 40%, and 0% of the points, and you can change them.
- Moodle's restrict access settings show an activity only when a grade condition, a completion condition, or a date is met. All conditions must be true before the activity appears.
- D2L Brightspace release conditions hide content, quizzes, and other items until a learner meets a condition, such as submitting an assignment or reaching a grade.
If your course already lives in an LMS, start there: build two paths after one quiz and see how many learners take each. For practice inside a video, Mindstamp exports a SCORM 1.2 or 2004 package so a compatible LMS records completion and score.
How to create an adaptive course or learning journey
You can build an adaptive learning path without an AI engine. One diagnostic question with a few routes covers the most common case, where some learners need review and others are ready to move on. These steps work in an LMS, an authoring tool, or interactive video.
1. Pick one objective and the mistakes learners make
Write the skill as something a learner can do, then list the wrong answers you see most often. Each common mistake becomes a route. If you can't name the mistakes yet, ask the people who teach or supervise the skill, or read last year's quiz results.
2. Write a diagnostic question that tells the mistakes apart
Each wrong option should map to one misconception. In the warehouse demo above, the two wrong shelves are wrong for different reasons, so one opens a refresher on the load rule and the other opens a side-by-side comparison.
3. Keep each route short and specific
A route should teach the one thing the learner missed and then rejoin the main lesson. A long remedial module turns a quick fix into a detour.
4. Check again after the route
End each route with a follow-up question at the same level as the original. Send a learner who misses it twice to a different explanation or to a person.
5. Decide when the learner moves on
Set the exit rule before you build. It can be one correct follow-up, a score threshold, or a mastery level. In Mindstamp, Repeat Until Correct shows Try Again after a wrong answer and asks the same question without revealing the correct answer. On supported plans, the running score in the cScore variable can drive conditional logic, for example sending a learner back when the score is under your pass mark (passing scores and retries).
6. Read the paths, then fix the routes
After the first group of learners, look at which route each person took and how many passed the follow-up. If most learners take the same review route, the main lesson is unclear. If few pass the follow-up after a route, that route needs a new explanation.
The same pattern works for a soft skill. In the next demo, a frustrated customer calls about a repeat problem, and your first reply decides which teaching scene you see.
To build this in Mindstamp, add the choices as buttons or a question, then send each answer to its scene with video branching. Adaptive routes work best when the learner does something at each step, so the ideas in our active learning examples and active learning strategies make good route content. For journeys that span several courses or a whole role, see our guide to personalized learning pathways.
Does adaptive learning work?
The research is positive but uneven. Three reviews show the range.
- Kulik and Fletcher's 2016 meta-analysis of 50 controlled evaluations found that intelligent tutoring systems raised test scores by a median of 0.66 standard deviations, roughly from the 50th to the 75th percentile. Gains were larger on locally developed tests than on standardized tests.
- SRI Education's 2016 evaluation of adaptive courseware covered more than 19,500 students at 14 colleges and universities. In 4 of the 15 implementations with usable data, average grades rose slightly. Most showed no difference in grades, and the odds of completing a course did not change. Moving from lecture to blended adaptive instruction did improve posttest scores (SRI report).
- A 2019 meta-analysis in BMJ Open pooled 21 studies with 3,684 health professionals and students. It found a large effect of adaptive e-learning on skills (effect size 1.19). The effect on knowledge (0.70) had a confidence interval that included zero, and the authors rated the risk of bias as generally high (Fontaine et al., 2019).
Our reading of this evidence: adaptivity helps most when a course is redesigned around it and when learners practice a skill. It does not fix completion rates by itself, so plan a human follow-up for the learners the system cannot move.
How to choose an adaptive learning platform
Ask these questions in every demo and pilot.
- Does it come with content for your subject, or do you write the content? Publisher courseware ships with its own question banks. Rule-based tools and interactive video use your material, so the real cost is writing every route.
- What evidence does it adapt on? Answers alone, or confidence, time, and problem-solving strategy too?
- Can a teacher or manager see why a learner was routed somewhere, and override it?
- Where does learner data go? US schools should ask how the vendor protects student records under FERPA and, for children under 13, how it complies with the FTC's COPPA rule.
- Does it connect to your LMS and send scores back?
- What will the pilot measure? Pick one course, one group of learners, and a comparison group before you start.
Plan for instructor buy-in as well. Show teachers or trainers the reports before launch and agree on what they will do with them.
Common questions about adaptive learning
Which of the following is an example of an adaptive learning platform?
The adaptive option is the one that changes the next question or lesson based on a learner's answers. ALEKS, DreamBox Math, and Duolingo all qualify, and any of them also answers the related question about which educational platform is an example of personalized adaptive learning. A video course where everyone watches the same lessons, a PDF workbook, or an LMS that releases identical content to every student does not adapt.
What is the difference between adaptive learning and personalized learning?
Personalized learning is the broad goal of fitting instruction to each learner's needs, pace, or interests. It can include learner choice and teacher planning. Adaptive learning is one way to get there: software changes the path automatically from the learner's responses.
How big is the adaptive learning market?
Estimates vary widely because research firms define the market differently. IMARC Group puts it at USD 5.8 billion in 2025 and forecasts USD 29.7 billion by 2034. Expert Market Research puts 2025 at USD 4.64 billion and forecasts USD 37.99 billion by 2035. OG Analysis sells a third adaptive learning outlook through MarketResearch.com. Treat any single figure as one firm's model.
What happened to Smart Sparrow?
Pearson bought Smart Sparrow's technology for USD 25 million in January 2020 to build into its own learning platform. EdSurge reported at the time that the original Smart Sparrow platform relied on Adobe Flash, which ended in December 2020, and that a small Smart Sparrow team would help customers move their content. If you used Smart Sparrow for branching lessons, the rule-based tools above cover the same pattern.
The fastest way to learn what adaptive design asks of you is to build one route. Take a video you already use, find the question learners most often get wrong, and give each wrong answer its own short explanation. You can try it in Mindstamp with a 14-day free trial that needs no payment card.
