← Back to the wire

Ask HN: Can you still tell AI-generated text apart in your own language?

SpeculationModelAug 17, 2026

The author reports that AI-generated Japanese remains detectable through recurring word choices, unlike AI-generated English, which has become harder to identify. The author notes that model choice matters more in non-English languages. Anthropic's Opus 4.6 and Fable 5 produce reasonably good long-form Japanese outside coding sessions, while Google's models handle a wider range of registers. OpenAI's models, Opus 4.8/5, and Fable in coding sessions sound overly geeky in Japanese.

Receipt № 14331 source · awaiting confirmation ◐

Evidence

1source· awaiting independent confirmation

No score is assigned. Sources and their independence are shown in the citation chain below.

Citation chain · 1 source

01medPRIMARY
OpenAICompanyGoogleCompanyAnthropicCompanyFable 5ModelFableCompanyOpus 4.6ModelOpus 4.8/5Model
Canonical: https://news.ycombinator.com/item?id=49338865