Most language-learning advice online is confident and unsourced. This site is built the other way round: an answer only exists if published research backs it.
Academic search engines don't respond well to conversational questions, so we first translate what you asked into several targeted queries. If you picked a language, some queries target that language specifically and others stay general, so we catch both.
We query Semantic Scholar, CrossRef, and PubMed/ERIC in parallel, then de-duplicate and rank what comes back by relevance, citation count, and recency. Where a paper has a legally free full-text copy — via Unpaywall, PubMed Central, or the author's own site — we read the full paper rather than just the abstract.
The AI is given the retrieved sources and instructed to use nothing else — no general knowledge, no filling in gaps. Every claim has to cite a specific source. We then check each citation programmatically and discard any that points at something we didn't actually supply.
If the research we found doesn't really answer your question, the report says so and explains what's missing, rather than dressing up thin evidence as a confident recommendation. The same applies per language: where a language doesn't have enough acquisition research indexed, we mark it as not ready instead of guessing.
It isn't a replacement for a teacher, and it isn't infallible. Retrieval can miss the best paper on a topic, and a summary can misread a study. The citations are there so you can check — please do.