Chotto triples user engagement by switching to Japanese-first ShisaLLM
Moving from a general-purpose model to Japan-tuned ShisaLLM drove daily sessions, retention, and paid conversion sharply higher.
Chotto.chat is Shisa.ai's Japanese-first consumer AI assistant. At launch, the product ran on a general-purpose large language model, and user feedback kept flagging the same issues: awkward honorifics, weak handling of colloquial and dialect speech, and surprising gaps in Japan-specific context (government paperwork, restaurants, gaming).
In late 2025, the Chotto team cut over the entire backend to our in-house ShisaLLM V2 family. We folded real, PII-scrubbed user conversation logs into the fine-tuning mix, with a specific focus on automatic register-switching between polite and humble forms and on handling of uniquely Japanese expressions.
Sixty days post-migration, messages per daily active user were 2.1× higher, 30-day retention was 1.8× higher, and conversion to the paid tier rose 3.0×. Gains were most pronounced with users aged 50 and above — a segment that previously had our worst drop-off — who consistently describe the new experience as "natural Japanese that never breaks honorifics."
Chotto is now piloting voice in/out with Shisa Speech API for a subset of users, and a mobile app is slated for general release during 2026.
More case studies
Leading Japanese media group cuts localization time in half with Shisa Translation API
Replaced a multi-vendor translation memory workflow with Shisa Translation API plus a custom glossary, halving weekly article localization turnaround.
Read case studyMajor JP retailer cuts contact-center running costs 40% with Shisa ASR API
Swapping a global cloud ASR vendor for Shisa ASR API plus automated call summarization dramatically reduced after-call work.
Read case studyYour team could be the next case study
Translation, speech, or a custom model — tell us what you're building and we'll help you pick the right mix of Shisa.AI products for the job.