Retrieval Strategy: How to Make Learners Remember.

A retrieval strategy is a plan to get people to pull knowledge out of their heads. They do not just read it again. Quizzes, flashcards, and short recall tasks are the usual tools.

If you teach for a living, this is one of the best-tested ideas in learning science. It is cheap. It is simple. And most courses still skip it.

This guide shows what it is, why it works, and how to build it into your own program. It also shows where the research is thin, so you do not oversell it.

TL;DR

• Retrieval means recalling something from memory. Rereading does not count.
• It works best when it is spaced out and followed by feedback.
• Short, low-stakes questions beat one big exam at the end.
• The research is strong, but not perfect. Most studies come from a narrow slice of the world.
• AI can run the questions and spacing for you, if your expertise is encoded properly first.

What a retrieval strategy actually is

Think of a kitchen. Reading a recipe again and again does not make you a cook. Cooking the dish from memory does.

Retrieval practice works the same way. The learner actively recalls previously learned information from memory instead of rereading it. Carnegie Mellon's teaching centre lists quiz questions, flashcards, practice problems, and recalling a concept during class as examples.

The focus is on getting information out, not putting it in. That one switch is the whole idea.

Retrieval practice vs retrieval strategy

People use these terms loosely. Here is a clean way to split them.

• Retrieval practice is a single act. One quiz. One brain dump.
• Retrieval strategy is the plan. What you ask, when you ask it, and what happens after.

A good plan covers four things:

1. What the learner must remember
2. How they will be asked
3. When the questions come back
4. How feedback arrives

We will cover each one below.

It does not have to look like a test

Many people hear "retrieval" and think of exams. That is too narrow. A recent review explains that retrieval activities need not look like conventional tests as long as they make the learner recall knowledge.

A prompt like "Write down what we covered last week" counts. So does a call where you ask a client to explain the method back to you.

Why it works

When you recall something, the brain rebuilds it. That effort is the point. Retrieval activates brain pathways that bring information back into working memory. Memory and understanding get stronger each time.

MIT's Teaching and Learning Lab says the same in plain terms. Actively retrieving information improves long-term retention compared with restudying it.

Hard on purpose

Here is the odd part. Learning feels slower this way. Karpicke's 2025 review calls these desirable difficulties. They slow early progress. They improve long-term retention and transfer (using knowledge in a new situation).

That is why learners often dislike it. Rereading feels smooth. Recall feels hard. The smooth way is the weaker way.

One review in Nature Reviews Psychology suggests effective strategies are underused partly because of false beliefs about learning and their counter-intuitive nature. Expect some pushback from your own students. Explain why it feels hard.

Even wrong answers help

This surprises people. Retrieval can help even when a student tries to recall and fails, as long as they get feedback afterwards.

So a miss is not a failure. It is data.

Side benefits

Karpicke's review also lists knock-on effects:

• Students see more clearly what they know
• Teachers spot gaps in what was taught
• Students feel more motivated to study

That second point matters if you run a paid program. Retrieval questions tell you where your own teaching is weak. Those mediated effects are free feedback on your course.

There is also a forward effect. Retrieving old material can improve the learning of new material that comes after it.

Techniques you can use this week

You do not need software. Start with these.

Quick, low-prep methods

• Brain dump. Learners write everything they remember on a topic. No notes allowed.
• Last-session recall. Open with: "What did we cover last time?"
• Two takeaways. End a session by asking for two things they learned. The retrieval practice site lists brain dumps, lecture recall, and two-takeaway endings as simple options.
• Flashcards. Question on one side. Answer on the other. Try to answer before you flip.
• Copy, cover, check. Write a fact. Cover it. Write it from memory. Check. UC San Diego's psychology department lists copy-cover-and-check and flashcards among its methods.

Methods for teachers and course creators

Carnegie Mellon suggests short-answer quizzes, lecture or reading recall, and peer teaching. Treat them as learning moments, not grading tools.

Here is how that looks in a paid program:

• Open every module with three questions from the last one
• Add a "teach it back" task in each community thread
• End each week with a one-page recall sheet
• Send a short question by email two days after a lesson

None of these need marking. That is a real plus. Educators can boost long-term learning without extra grading by asking students to retrieve instead of review.

Open or multiple choice?

Both work. A review of testing as a learning tool found open-ended and multiple-choice questions both gave learning benefits, and that retrieval beat simple restudy for long-term retention.

So pick what fits your setting. Multiple choice is fast to run. Open questions take more effort but show more of what a learner really thinks.

Space it out

One burst of practice is not enough. Spread it.

Research reviewed in PMC finds that spaced retrieval, spread over time rather than packed into one day, is consistently more effective for long-term retention.

The same source gives a rule of thumb. Spacing gaps of 10% to 20% of the time you want people to remember are suggested. The maths:

Need to remember for	Gap between practice (10% to 20%)
30 days	3 to 6 days
90 days	9 to 18 days
180 days	18 to 36 days

This is a guide, not a law. Treat it as a starting point and adjust.

The EEF, a UK education charity, makes a related point. Retrieval should not be a single, separate activity and knowledge needs to be recalled repeatedly. So do not bolt one quiz onto the end of a course and call it done.

A simple spacing plan for a 90-day course

1. Day of the lesson: a quick recall question at the end.
2. Two to three days later: a short question by email or app.
3. One to two weeks later: mix it into a new lesson.
4. End of the course: a final recall task that spans everything.

Keep each one under five minutes. Short and often beats long and rare.

Add feedback

Recall without feedback leaves errors in place. Always show the right answer.

Feedback does more than fix mistakes. The same PMC review says feedback enhances learning, and delayed feedback can sometimes strengthen memory. Learners are prompted to retrieve their original thinking.

In practice:

• Show the correct answer right after a quiz for fast fixes
• For deeper topics, try a short delay before you reveal it
• Ask the learner to say why they got it wrong

Keep the stakes low. If a miss costs a grade, people stop trying.

What the research does not say

We like retrieval practice. But we should be straight about its limits.

The good news. A review of 50 classroom experiments found medium or large benefits in 57% of effect sizes, across levels and subjects. McDermott's 2021 review says practising retrieval soon after learning can slow forgetting across many material types, ages, and abilities.

The caveats.

• In that same review, only 6% of experiments were done in non-WEIRD countries (outside Western, educated, industrialised, rich, democratic settings). Your global learners may differ.
• Against other methods, such as concept mapping, the benefits are not yet fully conclusive.
• Most studies are on school and university students. Adult professionals paying for an expert program are a different group.

That last point is our reading, not a finding. Test it with your own learners and watch the results.

How to build a retrieval strategy, step by step

Here is a build you can finish in an afternoon.

Step 1. List what must stick

Pick the 10 to 20 ideas a learner cannot work without. Not the whole course. Just the core.

For a founder with a method, this is the framework. The steps. The rules for when to do what.

Step 2. Turn each one into a question

Write a prompt that needs real recall.

• Weak: "Do you remember the pricing framework?"
• Strong: "Name the three inputs you check before you set a price."

Step 3. Put them on a calendar

Use the spacing table above. Each idea shows up at least three times.

Step 4. Write the feedback

For every question, write a short model answer. Add one line on the usual mistake.

Step 5. Read the results

If half your learners miss one question, the lesson is the problem. Fix the lesson.

Step 6. Repeat

Retire what everyone knows. Add what they keep missing.

Our view: your method is the real material

We work with educators and consultants who have built something valuable. A framework. A method. A way of making decisions that took years to learn.

Teaching it well once is the easy part. The hard part is that it fades. Learners finish a program, feel good, and the detail slips within weeks. The founder gets pulled back in to answer the same questions again.

That is a retrieval problem as much as a teaching problem.

Our view is simple:

• Your IP is not a pile of content. It is a set of decisions.
• Learners need to practise those decisions, not just read about them.
• The practice needs to happen when you are not in the room.

This is where we build. We take how a founder thinks and encode it into a network of AI agents (software helpers that each do one job). One can ask the questions. One can mark answers the way you would. One can nudge a learner on the right day. We build these with Claude Code, from your own method.

We are also honest about the limits. AI does not replace good course design. If the questions are poor, a faster system just delivers poor questions faster. Get the strategy right first. Then automate it.

For more on this thinking, see our guide on how to create an AI-powered learning experience and our piece on the role of IP in online education.

James Killick

Where AI fits in a retrieval strategy

Think of hiring staff. You would not give a new hire a binder and walk away. You would check they remembered it. You would ask questions. You would correct them.

AI can be that checker. It can:

• Write fresh questions from your lessons
• Send them on a spaced schedule
• Mark open answers against your standard
• Flag topics where many learners struggle

There are risks. The AI may mark an answer wrongly. It may invent a fact. So keep a human in the loop for anything that matters, and review a sample of its marking each week.

Memory matters for the AI too. An agent needs to know what to keep and what to drop. Our note on AI agent memory covers that. If you deliver content online, our guide to AI tools for course content delivery is a useful next read.

And once your learning content works, people need to find it. Tools built for AI-driven search visibility can help your pages show up in search and AI answers.

Want us to build it with you?

If you earn $1M or more from your expertise, and you are tired of being the bottleneck, we can help.

Our 90-day program is done with you. We pull your method out of your head. We turn it into a working system of AI agents. Learner practice and support is one of the jobs it can cover. You own what we build, with no vendor lock-in.

Here is what it costs, as published:

• Foundation Month: $3,500, one-off
• 90-day program: $18,500, one-off
• Platform Access: $90 a month or $900 a year, for a self-serve tool

See the full detail on our AI consulting services page. If you teach online, the page for online educators is the best fit. If you want to try the tool first, look at Platform Access.

Not sure your IP is ready? Our guide to packaging educator IP into digital products is a good place to start.

Common mistakes

Watch for these.

• One big quiz at the end. It tests. It does not train.
• High stakes. Fear makes people avoid recall.
• No feedback. Wrong answers stay wrong.
• Only easy questions. If it feels easy, it is not doing much.
• Never revisiting. Spacing is the other half of the plan.
• Skipping the explanation. Tell learners why you are asking. Otherwise it feels like busywork.

Next step

Want a retrieval strategy that runs without you? Book a call about our AI consulting services and we will map your method together.

Sources

• Retrieval Practice for Improved Learning, Eberly Center, Carnegie Mellon
• Help Students Retain, Organise and Integrate Knowledge, MIT Teaching + Learning Lab
• Retrieval-based learning, Karpicke (2025)
• Retrieval potentiates new learning, Chan et al. (2018)
• Using Testing as a Learning Tool, PMC
• Retrieval Practice Consistently Benefits Student Learning, Agarwal et al. (2021)
• Retrieval Practice in Classroom Settings, Frontiers in Education (2019)
• Practicing Retrieval Facilitates Learning, Annual Reviews (2021)
• The science of effective learning with spacing and retrieval practice, Nature Reviews Psychology (2022)
• Retrieval Practice, UC San Diego Psychology
• What is retrieval practice?
• Not another quiz! Refining retrieval practice, EEF