You know the card. You have seen it eleven times. You got it wrong ten of them. Every time it shows up, you stare at it, guess, flip it, and feel a little worse about your memory. It eats your review time and teaches you nothing.
Anki has a name for that card: a leech. And Anki is trying to tell you something when it flags one. A leech is not a sign that you are bad at memorizing. It is a sign that the card itself is broken. Fix the card and the forgetting usually stops.
What a Leech Actually Is
Every time you press Again on a card that is already in review mode, Anki counts it as a lapse. Rack up enough lapses on one card and Anki decides the card is draining more time than it is worth. By default, that cutoff is 8 lapses. At that point Anki tags the note with leech and, depending on your settings, suspends the card so it stops showing up in reviews.
That is the whole mechanism. There is no judgment in it. Anki is just keeping score: this card has failed 8 times, so something about it is not working.
To see your leeches, open the Browse window and search for tag:leech. If you have been studying for a while and have never looked, expect a pile. A third-year student working through a 25,000-card deck can easily accumulate 100 to 300 leeches without noticing, because each one only costs a few extra seconds a day. Added together, they can quietly eat 10 to 20 minutes of every study session.
Why Cards Become Leeches
A card does not become a leech because the fact is hard. It becomes a leech because of how the card is written. In practice, almost every leech comes down to one of five problems.
1. The card asks for too much
This is the most common cause by far. A card that reads "List the causes, presentation, diagnosis, and treatment of hypercalcemia" is really four cards pretending to be one. You might remember three of the four parts and still press Again, because you could not produce all of it. The card fails you even though you knew most of the answer.
The rule is simple: one card, one fact. If the back of the card needs a paragraph, the card is overloaded. Split it.
2. The prompt is vague
"What is important about gentamicin?" is not a question. It has no single right answer, so your brain cannot form a stable retrieval path to it. You might answer "nephrotoxicity" one day and "ototoxicity" the next, and both feel wrong because the card seems to want something else.
A good prompt names exactly what it wants: "Gentamicin causes which two toxicities that both end in -toxicity?" That card has one answer and you either know it or you do not.
3. Two cards interfere with each other
These are the evil twins. You have one card asking about loop diuretics and one about thiazides, and every time one appears you answer with the other drug's side effects. Or you keep mixing up Addison's disease and Cushing's syndrome because both cards ask about "the patient with an adrenal disorder."
Interference leeches feel uniquely frustrating because you do know the material. You just cannot keep the two facts apart under time pressure. The fix is not more reps. The fix is rewriting the cards so each one forces you to notice the distinguishing feature.
4. You never understood it in the first place
Some leeches are cards you made from a lecture slide or a premade deck without ever understanding the underlying idea. You memorized the shape of the answer without the reasoning behind it. Those cards fail forever, because there is nothing for the memory to attach to.
The classic test: if someone asked you to explain the answer out loud in two sentences, could you? If not, no amount of repetition will save the card. Go read the source material, understand it, then rewrite the card.
5. The detail is not worth remembering
Not every leech deserves to be saved. Some cards test professor-specific minutiae, a number from one slide of one lecture, or a fact so obscure that it will never appear on a board exam or a ward. If a card has lapsed 8 times and you would not miss it if it vanished, that is information. Delete it.
How to Fix a Leech: The Five Moves
When you sit down to clear leeches, every card gets one of five treatments. Decide quickly and move on. The goal is a cleaner deck, not a perfect one.
Split it
Take an overloaded card and break it into two to four narrow cards. The hypercalcemia card becomes:
- "What is the most common cause of hypercalcemia in outpatients?" (Primary hyperparathyroidism)
- "What is the most common cause of hypercalcemia in inpatients?" (Malignancy)
- "Hypercalcemia causes which ECG change?" (Shortened QT interval)
- "What is the first-line treatment for severe hypercalcemia?" (IV fluids, then loop diuretics once volume replete)
Each new card tests one retrieval. Each one is easy to grade as right or wrong. The four cards together take less total review time than the one monster card did, because you stop failing them.
Rewrite the prompt
Make the question specific enough that there is exactly one correct answer. Replace "What is important about X?" with "What is the mechanism / side effect / next step / most common cause?" Pick one. If you genuinely need three facts about a drug, that is three cards, not one card with a vague prompt.
Also check the card from the other direction: read only the front and try to guess what the back says. If you cannot tell what the card wants without already knowing the answer, the prompt is the problem.
Add a hook
A card with no context is a card with nothing to grab onto. Add the thing that makes the fact memorable: a one-sentence clinical vignette, an image, or the reasoning behind the fact.
Compare these two cards:
- Front: "What is the treatment for croup?" Back: "Dexamethasone."
- Front: "A 2-year-old has a barking cough, inspiratory stridor, and a steeple sign on neck X-ray. What is the diagnosis and first-line treatment?" Back: "Croup. Dexamethasone, plus nebulized epinephrine if severe."
The second card gives your brain a scene to attach the fact to. For visual topics like histology or rashes, an image on the front does the same job. This is also where cards you made from lecture slides often go wrong: a slide bullet copied verbatim has no hook. Rewriting it as a patient scenario fixes that, and it is the same reason converting slides into cards works best when each card gets its own narrow question rather than inheriting a whole bullet list.
Separate the twins
When two cards interfere, rewrite both so the difference is the point of the card. Do not just add more detail. Force a comparison.
Instead of two cards that each ask "What are the side effects of [drug]?", write:
- "Loop diuretics and thiazides both cause hypokalemia. Which one also causes ototoxicity?" (Loop diuretics)
- "Which diuretic causes hypercalcemia, and which causes hypocalcemia?" (Thiazides cause hypercalcemia; loops cause hypocalcemia)
Now each card tests the exact distinction you kept mixing up. If the two facts still blur together, add a brief note on the back explaining why the difference exists. Loop diuretics block the NKCC2 transporter in the thick ascending limb, where calcium reabsorption depends on the same gradient, which is why they waste calcium while thiazides retain it. A one-line mechanism turns a coin flip into a fact you can reason out.
Delete or suspend it
Some leeches should not be fixed. Delete the card if the fact is low yield, duplicated by another card, or wrong. Suspend it if the fact matters but not now, for example a Step 1-only detail you are carrying into clerkships. Deleting a bad card is not quitting. It is the same triage you do with any study resource: stop spending time on material with no return.
A practical rule: if you have already rewritten a card once and it lapses again, delete it. Two strikes and the fact is telling you it does not want to live in flashcard form. Learn it from questions or reading instead.
Settings That Matter: Threshold and Leech Action
You can change how Anki handles leeches under Deck Options, in the Lapses section. Two settings matter.
Leech threshold is the number of lapses before Anki flags a card. The default is 8. Some students lower it to 4 or 6 so problem cards get caught earlier, before they have burned weeks of review time. That is reasonable if you actually triage your leeches regularly. If you never look at them, a lower threshold just suspends more cards silently, so keep the default until the weekly routine below is a habit.
Leech action is what Anki does when a card crosses the threshold. The options are Tag Only or Tag and Suspend.
- Tag Only keeps the card in your reviews and just adds the
leechtag. Use this if you triage leeches weekly and want them to keep appearing until you fix them. The risk is that unfixed leeches keep costing you time. - Tag and Suspend pulls the card out of reviews immediately. Use this if your backlog is large or you are in a busy rotation and cannot afford to keep re-failing the same cards. The risk is forgetting the suspended cards exist, so you still need the weekly triage to process them.
Neither setting fixes leeches on its own. The setting just decides how loudly Anki complains. The fix is always the rewrite.
Ease Hell and the Again Button
Leeches and ease hell are cousins. Every time you press Again on a review card, Anki lowers that card's ease factor, which shortens all of its future intervals. Press Again on the same card repeatedly and its ease can bottom out at 130 percent, meaning the card comes back constantly no matter how well you know it by then. That is ease hell: a card you have effectively learned but that Anki keeps showing you every day or two.
Two habits prevent it. First, be honest with the grading buttons. If you got the card right, even slowly, press Good, not Hard. Hard shrinks future intervals and pushes the card toward ease hell for no benefit. Second, if you truly did not know a card, do not just press Again and move on. Stop for ten seconds and read the back properly. Re-reading the answer once with attention is worth more than three additional failed reps.
If you use FSRS, the newer scheduling algorithm, ease works differently and the classic ease hell pattern is less of an issue. Leeches still exist under FSRS, though, because the leech counter tracks lapses, not ease. The diagnosis and the fixes in this post apply either way.
Leech or Just a Hard Card?
Not every card you fail is a leech. A hard card fails a few times, then stabilizes. You press Again twice in the first week, then Good from then on, and the interval grows. That is the system working as designed. Do not rewrite those cards. They are doing their job.
A leech is different in two ways. First, the failures never stop. The card lapses at a one-week interval, then again at a two-week interval, then again after a month. Second, the failures feel random. You get it right one day and wrong the next, with no sense of progress. That randomness is the tell. A hard card gets steadily easier. A leech stays a coin flip no matter how many times you see it.
If you are unsure, check the card's review history in the Browse window. Click the card, open Card Info, and look at the lapse count and the interval pattern. Two or three lapses with growing intervals is a hard card. Six or more lapses with intervals that keep collapsing is a leech, and it is time to rewrite.
One more distinction matters: a card can be hard because the fact is genuinely difficult, like the steps of the complement cascade, or because the card is badly written. Difficult facts reward splitting and hooks. Badly written cards reward rewriting the prompt. Either way, the answer is a better card, not more reps on the same one.
The Weekly Leech Triage
Leech management only works as a routine. Here is one that takes about 15 minutes a week.
- On the same day each week, open Browse and search
tag:leech. - Sort by lapse count so the worst cards are on top.
- For each card, pick one of the five moves: split, rewrite, add a hook, separate the twins, or delete/suspend. Spend at most one minute per card.
- After processing, remove the
leechtag from any card you fixed so it can re-enter reviews clean, and unsuspend any card you repaired. - Track the count. A healthy deck trends toward fewer leeches over time. If your leech count keeps growing, the problem is upstream: you are creating overloaded or vague cards faster than you are fixing them.
That last point is the real diagnostic. A handful of leeches is normal. Everyone has cards that do not click. A leech count that grows every week points upstream: you are creating overloaded or vague cards faster than you are fixing them. Slow down when you make cards, keep each one to a single fact, and write prompts that have one right answer.
The Bottom Line
A leech is Anki doing you a favor. It is pointing at the exact cards where your deck is wasting your time, and handing you a list of them. Most students ignore the list and keep paying the tax every day.
Spend fifteen minutes a week on it. Split the overloaded cards. Rewrite the vague prompts. Separate the twins. Delete the trivia. Your daily reviews will get shorter, your retention will go up, and the cards that remain will be the ones worth remembering.
And the next time a card fails for the eighth time, do not sigh and press Again. Open it up and ask what is wrong with the card. The answer is usually obvious, and the fix usually takes less than a minute.