Two loud camps argue about how adults should learn languages. The spaced-repetition camp says you memorise the highest-frequency vocabulary with SRS software and the rest takes care of itself. The immersion camp says you dump yourself into comprehensible input, ignore explicit study, and let the brain acquire the language the way a child would. Both are half right, and the mistake most learners make is treating them as alternatives instead of ingredients.
Here is the honest picture from the research and from what actually works for adults in the real world with 30 minutes a day.
What spaced repetition is actually good at
SRS is a memory tool, not a language-learning tool. It excels at one specific job: moving discrete facts from short-term memory into long-term memory with the minimum possible number of reviews. A well-tuned deck can lock in a hundred vocabulary items per week with about fifteen minutes of daily work.
What SRS cannot do is teach you to use those items in a sentence, in real time, with another human. It builds a lexicon, not a language. Learners who mistake it for the whole method end up with impressive Anki stats and an inability to hold a conversation. The lexicon is real, but it is passive — recognition, not production.
What immersion is actually good at
Immersion — meaning consistent exposure to comprehensible input in the target language, whether audio, video, or text — is what teaches the brain how the language patterns. Sentence rhythm, word order, register shifts, the difference between how people write and how they speak. None of this can be memorised as a fact; it has to be absorbed over hundreds of hours of exposure.
What immersion cannot do is efficiently deliver low-frequency vocabulary. If a word appears once every fifty hours of input, you will meet it about once a month at reasonable study volumes, and forget it between meetings. Immersion is a high-frequency delivery system; the long tail of vocabulary needs help.
The mix that actually works
The honest research consensus is that for adult learners, the ideal daily split at A1–B1 is roughly 20% SRS, 40% comprehensible input, 40% production. The exact numbers shift with level — beginners lean more toward input and drill, intermediates more toward production — but the general shape holds.
A 30-minute daily budget therefore breaks down as: six minutes of SRS in the morning with a small, well-tended deck; twelve minutes of listening or reading at slightly above your current level; twelve minutes of speaking practice, either with a human or an AI tutor. Not one of these three can be replaced by more of another.
Why most learners get the ratio wrong
The typical self-taught learner over-invests in one component. The SRS enthusiast has a 500-card deck and forty hours of accumulated review time but has never spoken a sentence out loud. The immersion enthusiast has watched two hundred hours of Netflix in the target language and cannot summon the word for "kitchen" without a five-second pause.
Both are missing the same thing: production. Production is the least fun component and the most tiring, which is why it is the one people avoid, which is why it is the one that separates plateaued learners from fluent ones. Whatever you cut, do not cut the twelve minutes of daily speaking.
What kind of SRS deck is worth building
A useful deck at A1–B1 has three properties. It is small enough to keep up with (no more than 500 mature cards). It contains only vocabulary you have already met in context (not raw wordlists — words you have heard or read in a sentence). And it uses cloze sentences rather than isolated words, so the retrieval is closer to production.
A card that looks like "kitchen — cocina" is nearly useless. A card that looks like "Voy a la ___ a preparar la cena" with "cocina" as the answer trains the word in exactly the retrieval shape you will need it in when you speak. This is the non-obvious difference between an SRS deck that transfers to speech and one that builds a museum of dead vocabulary.
The role of AI tutors in this mix
The last two years have collapsed the cost of the production slot from ten dollars for a human tutor session to essentially zero for an AI roleplay. This is the biggest practical change in adult language learning since Duolingo, and it has a specific effect: production practice, previously the bottleneck, is now the cheapest slot in the routine.
The right way to use this is not to abandon SRS or immersion — those still cover jobs the AI cannot — but to make production the daily anchor. See the AI tutor roadmap to B2 for the specific prompts and cadence. For why production matters so disproportionately, recognition is not production covers the mechanism.
The honest bottom line
Nobody has to pick a side. The learners who reach B1 in a year are running all three engines every day at the right ratio. The learners who plateau are almost always overweighting one and neglecting production. Fix the ratio and the same daily twenty-five minutes start compounding again — often after months of no visible progress.
Frequently asked questions
Is Anki still the best spaced-repetition tool?
For raw efficiency, yes. For daily adherence over months, no — most learners abandon Anki within eight weeks because the deck maintenance is boring. A slightly less optimal SRS you actually keep using beats a perfect one you drop.
Can I skip flashcards entirely if I do enough immersion?
For high-frequency vocabulary, yes. For low-frequency and false-friend vocabulary, no — immersion buries these under noise. Fifteen minutes a day of targeted SRS covers what immersion cannot see.
How much daily immersion do I actually need?
At A1, twenty minutes of comprehensible input a day is plenty. At B1, thirty to forty-five. Beyond that, diminishing returns set in unless you also increase output.
Is watching Netflix with subtitles really immersion?
Only if the subtitles are in the target language, not your native one. Native-language subtitles let your brain switch off the target audio entirely and effectively give you zero language input.