Science-backed answers to your language-learning questions

It's built specifically to avoid that. Every claim in a report has to cite one of the real sources we retrieved for your question — nothing is answered from the model's general training knowledge. After generation, we programmatically check every citation and strip out any claim that doesn't actually point at a source we supplied. If we can't back something with a real citation, it doesn't appear in your report.

Published, indexed academic research — then read the full text where a legal open-access copy exists, not just the abstract. Every source is linked in the report so you can check it yourself.

A general chatbot answers from whatever it absorbed during training — a mix of good research, forum opinions, and confident-sounding guesses, with no way to tell them apart. We only let the model use sources we've just retrieved and verified for your specific question, and we show you exactly which one backs each claim.

You can ask your first question with no account at all. Signing up gives you a small free allowance, and paid plans raise that limit for people asking regularly — every tier gets the same fully-cited reports.

We tell you, instead of dressing up thin evidence as a confident answer. Some languages and topics have a much deeper research base than others — where the evidence is limited, the report says so explicitly rather than guessing.

33+ languages, from deeply-researched ones like Spanish and Japanese to early beta coverage for languages with a thinner published evidence base. You can also ask general questions about learning that aren't tied to one language.