How to choose a study app: a ten-point checklist you can run in ten minutes

Choose a study app by what it makes you do, not by its feature list. It should make you answer before it shows the answer, bring ideas back on a spaced schedule, check your work against something verified, show where its content came from, say when it does not know, follow your teacher's rule, let you export everything, and tell you where your data lives. Ten checks below, each runnable in about a minute, with a worked run-through on a chemistry unit and the checklist as a table you can copy.

Why "best study apps" lists do not help you

A list of the best study apps ranks features and, often, sponsors. Dark mode, streaks, a pretty graph, a chatbot. None of those is a reason an app would raise a mark.

The research ranks something else: behaviours. Dunlosky and colleagues (2013) reviewed ten common study techniques against the evidence. Practice testing (recalling or solving before you look) and spacing study over days came out with high utility. Rereading and summarising came out low. An app is good if it produces the high-utility behaviours in you and bad if it produces the low ones, however polished it looks. A beautiful app that shows you the answer first and then asks you to rate yourself has built rereading with a nicer font.

This page names no product, on purpose. A ranking is out of date the month after it is written and only covers the apps the writer tried. A checklist applies to every study app, including the ones that do not exist yet, and you can run it yourself in ten minutes on whatever you are considering. Chapter 7 of the free playbook is the same idea on one page: use AI and apps that teach, and the test for which ones do.

The ten checks: how to choose a study app in ten minutes

Each check has a one-minute test and what a pass looks like. Run them in order on the free tier, before you move a term's notes in.

1. Self-testing first. Open a card or a practice question. Can you see the answer before you have tried? A pass is no: the attempt is required, or the app defaults to hiding the answer until you type or say something. Why it is first: in the standard experiment, students who recalled a passage remembered more a week later than students who reread it, and the rereaders felt more confident while remembering less (Roediger and Karpicke 2006). The full method is in the active recall guide.
2. Spacing. Make two cards and look for when they are due. Does the app bring an idea back on later days, with the gap growing from how you did? A pass is a schedule you can see and an algorithm the app names. Study spread over separate days beat the same time in one block across hundreds of comparisons (Cepeda et al. 2006). The spacing guide explains what a good schedule looks like.
3. Checked answers. Type a wrong answer to a maths or science question. Is it marked wrong against a verified solution, or does the app ask you to rate yourself on everything? A pass is a checker for computable answers, with self-rating only where nothing can be checked, such as an essay plan. Self-rating is honest in theory and generous in practice. The guide to solver apps has the longer argument.
4. Sources. Ask where an explanation came from, or look for a citation under it. Does every explanation say where it came from, and can you open it? A pass is a citation to your own material or to a licensed text, not "studies show". This is one of the four-part test for any AI study tool.
5. "I don't know". Ask the app about something you did not give it: "What does page 300 of my textbook say?" when you uploaded nothing. A pass says plainly that it was not given that page. A fail answers confidently about a page it has never seen, and if it will do that, it will do it when you cannot tell. The AI tutor guide shows what an honest refusal looks like.
6. Your own material. Can the app work from your lecture recording, your PDF, a photo of the board or a video link, and can you see what it did with them? A pass turns your material into notes or cards and lets you trace each one back. The guide to notes from a lecture or a PDF is what to do with the output.
7. The teacher's rule. Can your course's rule about AI help be set and followed? A pass is a setting the teacher or you control, and an app that declines to hand over an answer where the rule says no, and teaches instead. Whether using AI to study is cheating depends on that rule, and the guide to that question goes course by course.
8. Export. Can you leave with everything, cards, notes and schedule, in an open format? A pass is CSV, Markdown or the .apkg deck format, and you test it in the first week with a small deck. Cards you cannot export are rented. The flashcard maker guide covers importing and exporting in detail.
9. The free tier. Read what free includes. Then read what happens when it runs out mid-week, and whether there is a hard paywall at exam time. In our own reading of 320 one- and two-star reviews of study and answer apps (October 2026), the largest complaint about study apps was a paywall that appeared at crunch time, when the student had the least room to switch. A pass is a free tier that says its limits plainly and still lets you study and export when the limit is reached.
10. Where your data lives. Does the app say where your notes and recordings are stored and who processes them? A pass is a plain-language answer naming the country, saying whether your words leave it, and a privacy policy you can read in five minutes. Your recordings are also recordings of your teacher and your classmates, which is a second reason to ask.

The three red flags

Three things that should end the trial early, each for a reason the research gives.

A streak or leaderboard that rewards opening the app. A streak counts days you showed up, not ideas you recalled. If the app's main motivation loop can be kept alive by opening it and flipping three easy cards, it is training you to open the app, which is not the same as training you to remember chemistry. Rewards tied to recall attempts, or to nothing, are the better sign.

A "summary" feature as the main event. If the first thing the app offers to do with your notes is summarise them, it is leading with one of the two low-utility techniques in Dunlosky's review. A summary written by a model is a summary you did not write, read once. It can be a skim; it is not studying.

Answers with no source and no way to check. An app that produces an answer, cites nothing and marks nothing is asking you to trust it. In the same review bank, "the answer was wrong and I trusted it" was the largest complaint theme about answer apps. There is a research reason to care beyond the reviews: Bastani and colleagues (2025) gave high-school mathematics students an unrestricted AI assistant for practice and found that, when it was taken away for the exam, they scored about 17 per cent lower than students who had never had it. A version of the assistant that coached without handing over answers largely removed the harm. Check 1 exists because of that result.

The ten-minute run-through: a worked example

A Grade 12 chemistry (SCH4U) student is in the equilibrium unit, the sixteenth and seventeenth ideas in the course: the equilibrium constant, then Le Chatelier's principle. The student opens an app, any app, and spends one minute on each check.

Minute 1, self-testing. Types "What does a large value of K tell you about an equilibrium?" Did the app hint and ask for a try, or answer at once? Written down: hint or answer.

Minute 2, spacing. Makes two cards from the unit and looks for a due date on each. Tomorrow? Three days? Is the algorithm named anywhere?

Minute 3, checked answers. The student enters a wrong answer on purpose. For N₂ + 3H₂ ⇌ 2NH₃ with the equilibrium concentrations [N₂] = 0.50 mol/L, [H₂] = 1.0 mol/L and [NH₃] = 2.0 mol/L, the student writes K = 4.0. The right value: K = [NH₃]² / ([N₂][H₂]³) = (2.0)² / (0.50 × 1.0³) = 4.0 / 0.50 = 8.0. The wrong answer of 4.0 is what you get if you forget the denominator entirely, which is a common slip. Does the app mark 4.0 wrong against 8.0, and does it point at the missing denominator, or does it ask "How confident are you?"

Minute 4, sources. Asks where the definition of the equilibrium constant came from. A page in a licensed text, the student's own notes, or "studies show"?

Minute 5, "I don't know". Asks "What does page 300 of my textbook say about Le Chatelier's principle?" with no textbook uploaded. A plain "I was not given that page" is the pass.

Minute 6, own material. Uploads one lecture PDF from the unit and looks for what the app made of it, and whether each note traces back to a page.

Minute 7, the teacher's rule. Looks for a setting for the course's rule on AI help. Is there one, and who controls it?

Minute 8, export. Exports the two cards from minute 2 and opens the file. Are the questions and answers there in plain text?

Minute 9, the free tier. Reads the limits. What runs out, when, and what is left when it does?

Minute 10, the data. Opens the privacy policy and looks for the storage country and the processors. Found in five minutes or not.

Score it out of ten. The point is the method, not a verdict on any one app. A student who does this once knows more about the app than any ranking could tell them, and knows it about the app as it is this month.

The checklist to copy

Here is the checklist as a table: copy it into your notes, or write the ten lines on an index card and keep it beside the laptop while you try an app.

#	Check	One-minute test	Pass
1	Self-testing first	Open a card: can you see the answer before trying?	No; the attempt is required or the default
2	Spacing	Make two cards; find the due dates	A visible schedule with a named algorithm
3	Checked answers	Enter a wrong maths answer on purpose	Marked wrong against a verified solution
4	Sources	Ask where an explanation came from	A citation you can open
5	"I don't know"	Ask about a page you never gave it	It says it was not given the page
6	Your own material	Upload one lecture PDF	Notes or cards you can trace back to it
7	The teacher's rule	Look for a course AI-help setting	A setting that is followed, declining where it must
8	Export	Export two cards; open the file	CSV, Markdown or the .apkg deck format, readable
9	The free tier	Read the limits and what happens at the limit	Plain limits; studying and export still work
10	Where your data lives	Read the privacy policy's storage lines	Country named, processors named, in five minutes

What to do once you have chosen

The first week decides whether the app becomes a habit or a folder of unused cards. Write the cards for the hardest ideas yourself, by hand, because the writing is a small act of learning that a generator skips, and the guide to writing flashcards in the study-methods section has the rules. Import the sets you already have and run the six checks from the flashcard-maker guide on a sample of them. Process every recording and PDF within a day, while you can still check the notes against your memory of the class. Then put the app into a weekly plan, with the study-plan guide in the how-to-study section as the template, so the spaced schedule has days to land on.

The ten checks on one screen

1. Self-testing first: the answer is hidden until you try.
2. Spacing: ideas come back on later days, on a schedule you can see.
3. Checked answers: maths is marked against a verified solution.
4. Sources: every explanation says where it came from.
5. "I don't know": it admits what it was not given.
6. Your own material: your lecture, PDF, photo or video becomes something traceable.
7. The teacher's rule: set, and followed.
8. Export: everything leaves with you, in an open format.
9. The free tier: plain limits, no paywall that strands you at exam time.
10. Where your data lives: the country and the processors, in five minutes.

Ten minutes, ten checks, and no ranking needed.