What the Internet Says
Observations gathered from language learning communities, forums, and discussions across the internet.
Better than nothing, but worse than almost any serious alternative.
Near-universal agreement that it's fine for motivation and basic vocabulary, but the gamification eventually becomes the goal rather than language learning. Frequently described as "good for building the habit of showing up, bad for learning a language." Plateaus hard at around A2.
Infrastructure, not an app. Also easy to over-engineer.
Beloved by AJATT/Refold learners, controversial elsewhere. The real criticism isn't that SRS doesn't work. It clearly does. The criticism is that many people spend more time setting up and managing their Anki decks than actually using the language. "Anki should take 10–20 minutes a day, not 2 hours" is a common piece of community advice.
Most valuable at the beginning and at the advanced level. Less so in between.
Most experienced learners report that explicit grammar study is most valuable at the beginning (to orient yourself) and at the advanced level (to refine precision). In the intermediate phase, more input and output tends to produce faster results than more grammar drilling.
Don't wait more than 3–6 months for most languages.
The "speak from day 1" vs. "wait until you have a foundation" debate has no community consensus, but there's growing agreement that waiting more than 3–6 months for most languages is probably too long. Crosstalk and comprehension-focused tutoring sessions reduce the pressure enough that most people can start earlier than they think.
"Apps are the fuel, not the engine."
The Reddit community broadly treats apps as a supplement, not a method. The exception is Anki, which is treated as infrastructure rather than an app. No app alone will take you to fluency. They're most useful when plugged into a real method with real input.
Works better for some languages than others.
The AJATT/Refold framework was developed for Japanese, which has an enormous amount of native content available online. For rarer languages, the "find native content at your level" advice breaks down fast. For Swahili, Mongolian, or Welsh, a more textbook-and-community-based approach often makes more sense.
Consistency beats optimization. Every time.
Every method discussed in this section works for learners who stick with it. The number one predictor of success in self-directed language learning is not method quality; it's whether the learner stays engaged long enough for the method to work. The "best" method for you is the one that maintains your interest.