Is Anki worth it? An honest verdict, with a disclosure of interest

Everyone who studies seriously has gotten the same recommendation: "use Anki." You downloaded it, opened it, and found an interface that hasn't changed much in years, a manual the size of an exam syllabus, and a universe of presets, note types, and decks that nobody explained. The question that brought you here is legitimate: does that effort pay off, or is there a better path? The short answer is in the three paragraphs below. The full accounting (what Anki gets right, what it charges, and how to decide in your case) follows.

Yes, Anki is worth it for a specific profile: a zero budget, the desire to control every parameter of your own studying, and a tolerance for building the system before using it. For that profile it's hard to beat: the desktop program is free and open source, AnkiDroid is free on Android, and the science it runs is the best-documented in the psychology of learning.

For everyone else (people who want to study today, not administer a study system), the verdict flips. Anki's real cost was never money: it's the queue of decisions it hands off to you. Which cards to create, how to write them, which of the dozens of deck options to tweak, which algorithm to turn on. Each of those decisions is a chance to quit before you reach the part that works: the review itself.

And there's a detail almost no review mentions: Anki's best algorithm ships turned off. The explanation is right below, and it sums up the whole product.

Disclosure of interest: this blog belongs to Sift, the study app we built, which applies the same science as Anki with the opposite philosophy. This verdict doesn't pretend to be neutral; it makes up for that with transparency: every number has a primary source linked, and the credit we give Anki in the next section is sincere.

What Anki gets right, and why it deserves the hype

First, the credit, because it's substantial. Anki took spaced repetition out of the lab and put it on any student's computer: an effect measured since 1885 became a free, open tool that is still under active maintenance today. Much of what we know about studying with flashcards has passed through it.

The pricing model is honest in a way that's rare in this market. The desktop program (Windows, Mac, and Linux) is free and open source, under the AGPL-3.0 license, currently on version 26.05. Syncing through AnkiWeb is free. AnkiDroid, on Android, is free and maintained by community contributors. The only charge in the ecosystem is AnkiMobile, on iPhone: US$24.99 as a one-time purchase. It's that official purchase that funds the development of everything else.

And the customization power is real, not a marketing promise: card type, formatting, media, daily limits, scheduler parameters: nearly everything is adjustable. For anyone who enjoys tuning the system until it fits like a glove, that's a feature, not a flaw.

How Anki works: the built-in science is solid

The mechanic is simple to describe: each card is a question, you try to answer from memory before flipping it, and the algorithm sets the next review date based on whether you get it right or wrong: short intervals for what slipped away, long ones for what stuck. That mundane gesture runs, in a single move, the only two techniques awarded a high-utility rating when Dunlosky et al. (2013) assessed ten study methods: practice testing and distributed practice. Rereading and highlighting (nearly every student's bread and butter) landed on the low-utility shelf, at least the way they're used in practice.

The numbers behind that rating are big. In the meta-analysis by Cepeda et al. (2006), which pooled 317 experiments, spaced study yielded 47.3% correct on the final test versus 36.7% for massed study (with the same total study time). And in the experiment by Roediger & Karpicke (2006), one week after studying, people who had read a text once and tested themselves three times recalled 61%; those who had reread it four times recalled 40%. The cruel detail: the rereading group was the most confident it would do well (4.8 versus 4.0 on a 7-point scale). And it recalled the least. That's the illusion of fluency in action, and it's why the discomfort of trying to recall (Bjork's "desirable difficulties") isn't a side effect of the method: it is the method. The full mechanism is in the active recall guide.

Even the urgency rests on data. In the most hostile scenario ever measured (nonsense syllables, no review at all), about two-thirds of the memorization effort was lost in the first 24 hours, a curve that a 2015 replication (Murre & Dros, PLOS ONE) confirmed 130 years after the original. (The "we forget 90% in a day" line going around is a myth; the honest math is already scary enough.) If that curve describes your routine, the full diagnosis is in why you forget what you study. The point of this section is just one: Anki's premise is solid. The discussion from here on is the price of using it.

Where Anki charges: the friction bill

The first installment is setup. Anki doesn't greet you with a path; it greets you with a dashboard. Note types, presets, new-card limits, learning steps: decisions that show up before your first review is even done. The desktop interface is stable and functional, but it carries layers of years of accumulated decisions: anyone used to modern apps feels the step-up right on the first screen.

The second installment is the inventory. In Anki, a good card is a card you write, and writing good questions about everything you study is a second job. Shared decks exist, but the curation is yours, and the review queue is only as good as the material feeding it.

The third installment is the most revealing. Anki's default scheduler is still the one based on SM-2, a scheme the official documentation now treats as legacy. Its successor exists inside the app itself: FSRS has been native since version 23.10, from October 2023, with no add-on required, and the release notes themselves state that it improves on SM-2's scheduling. It's the state of the art in its category: trained and evaluated on an open benchmark with data from 10,000 Anki users and about 727 million reviews (nearly 350 million used in the evaluation). But it ships turned off: you're the one who has to enable it in the deck options, preset by preset, with a default target retention of 90%. The perverse result: the user who would benefit most from the best scheduler, the beginner, is exactly the one who doesn't know it exists.

None of this is an accusation of bad faith; it's product philosophy. Anki maximizes control and assumes you'll push through the friction. Configuration solves much of the criticism above, except that each adjustment is one more toll, and tolls are exactly where study routines die.

Anki on the scale: a category-by-category comparison

The same tool, seen from both sides of the counter:

Category	What Anki delivers	What it asks in return
Price	Free on desktop and Android; US$24.99 on iPhone (one-time purchase)	Nothing. It's the strongest point
Science	Self-testing + spacing, Dunlosky's (2013) high-utility pair	Nothing. The foundation is solid
Algorithm	FSRS native since 2023, state of the art	Ships turned off; the default is still SM-2, treated as legacy
Cards	Full control of format, media, and content	You write (almost) everything from scratch
Configuration	Nearly everything is adjustable	Nearly everything has to be decided by you
Consistency	Rewards those who get through the setup	The friction charges before the method pays off

The question that decides: which app will you still be using in March?

To put it plainly: the right question isn't "which app has the most features," but "which one you'll still be using in March." A study tool isn't judged on its feature tour; it's judged in the tenth week, when the motivation is gone and only the friction is left. By that yardstick, Anki is unbeatable for a minority; for everyone else, it becomes the perfect system that got finished and sat there.

Sift is the opposite bet on the same science: if the evidence says that what builds memory is spaced self-testing, then everything standing between you and the review is the enemy. The cards are generated from the material you study, instead of written by hand; FSRS works out of the box, with no hidden option; the day's review arrives already scheduled, with no configuration decision along the way. Less of a control panel than Anki? No doubt. That's the trade. The bet is that consistency is worth more than customization.

What to do today: decide in 5 steps

1. Answer one question before choosing the app: when a system demands configuration, does it energize you or drain you? That answer is worth more than any review, including this one.
2. Anki profile (zero budget, joy in tuning): download the free desktop version and start with a small deck you wrote yourself: 20 cards about what you studied this week.
3. Turn on FSRS on day one, in the deck options: it ships turned off, and it's the biggest free improvement available inside the app.
4. "I want to study, not configure" profile: look for a tool that generates the cards and schedules the reviews for you. And spend the saved energy on the studying itself.
5. Whatever the choice, the method doesn't change: questions instead of rereading, spaced reviews instead of last-minute cramming. The app is the secretary; the method is the boss.

Frequently asked questions

Is Anki free?

On desktop (Windows, Mac, and Linux), yes: free and open source, with AnkiWeb sync also free. On Android, AnkiDroid is free and community-maintained. On iPhone, AnkiMobile costs US$24.99 as a one-time purchase. It's the official purchase that funds the project.

Is Anki good for a competitive civil-service exam?

Yes, high volume over a long timeline is the scenario where spaced repetition pays off most, and black-letter statutes turn into card questions naturally. The bottleneck is the discipline of creating and maintaining the cards alongside the exam syllabus; the app matters less than the plan around it, and that plan is in how to build a study plan for a competitive civil-service exam.

What is FSRS, and do I need to turn it on?

FSRS is the state-of-the-art spaced-repetition scheduler, evaluated on an open benchmark with hundreds of millions of real reviews. In Anki, it has been native since version 23.10 (October 2023), but it isn't the default: it ships turned off, and enabling it is manual, in the deck options, preset by preset.

Is there an alternative to Anki?

It depends on what bothers you. If it's only the algorithm, you don't even need to switch apps: turn on FSRS. If it's the friction (writing cards by hand, hunting for settings), the alternative is the category of apps that already ships the modern algorithm turned on and the schedule ready to go; Sift is our house bet in that category.

If you've read this far, the verdict is already yours: Anki rewards those who pay the friction toll, and it deserves every bit of praise it gets from those who paid it. Sift, which we built, makes the opposite bet: the same science, with friction as the enemy. In practice, three concrete differences from Anki:

• The state-of-the-art algorithm is the default, and it adapts to you. The scheduler is FSRS-6, and it optimizes itself for your individual memory: it relearns its own parameters from your history of hits and misses, instead of applying a generic curve. It's not a mode you have to enable preset by preset: it's the default.
• You don't start from an empty deck. You can adopt and evaluate ready-made community decks, or import your own, instead of typing every card from scratch.
• The date bookkeeping isn't yours. Each card enters the queue with its own forgetting curve, and the review hands each one back on the day answering it pays off most.

Where Anki asks for configuration patience and maintenance discipline, Sift takes that layer off your shoulders. What stays yours is the only part that builds memory: studying and answering.