Star Tribune AI translation investigation — Poynter case study

When a shooting occurred at a Minneapolis church in August 2025, the Minnesota Star Tribune’s AI Lab became a critical reporting asset. The shooter’s journal was written in Faux Cyrillic, a Russian typography that directly translates English words, and the newsroom needed translations on deadline.

Context

Dana Chiueh, then the Star Tribune’s news innovation engineer, took screenshots of the journal pages and uploaded them to ChatGPT. The first task was simply identifying the language. From there, a team of four journalists worked with AI to translate the entries, a process that was time-consuming because AI would sometimes generate incorrect or duplicate words due to illegible handwriting.

The translations totaled nearly 200 pages containing more than 150,000 words in Cyrillic and English. A human translator estimated this would take weeks; the AI-assisted process was completed in days. The findings provided leads including names of friends, family, and coworkers, and revealed the shooter’s violent and racist thoughts.

Key takeaway

This case represents AI at its most clearly valuable for journalism: handling a task that is too large and too urgent for manual processing, where human judgment is still applied to the outputs. The reporter who wrote the resulting profile described it as “a great use case for AI,” precisely because the technology handled translation volume while humans handled reporting, verification, and narrative.

Who should read this

Journalists, editors, and newsroom managers considering how AI can extend reporting capacity for investigative work, particularly in situations involving large volumes of unstructured text.