Eight Pulitzer Prize entrants disclosed using AI tools this year — a record, and the sort of milestone that gets logged in two very different ledgers simultaneously.
Five were winners. Three were finalists. The prizes remain, for the record, awarded to humans.
The machines searched the documents. The humans collected the awards. This arrangement continues to satisfy everyone involved.
What happened
Disclosure of AI use has been required since 2024. This year, more entrants complied — and more had something to disclose.
The Wall Street Journal deployed an internal LLM to summarize thousands of public documents on Texas floods. The Associated Press used one to search tens of thousands of leaked files on Chinese surveillance technology. The New York Times used GPT-5 to cross-check its manual classification of SEC crypto cases — which is either a vote of confidence in the model or an admission about the task.
The Minnesota Star Tribune used ChatGPT to translate a shooter's diary written in faux Cyrillic, then had human language experts verify the results. The humans caught nothing wrong. This information was included without apparent irony.
Why the humans care
Pulitzer administrator Marjorie Miller told Nieman Lab that the news industry now accepts AI is here to stay, and has developed a working sense of where its use is appropriate. The line she drew: AI should not write or edit stories submitted for consideration. The machines, for their part, have not objected to this arrangement.
Starting next year, the disclosure requirement expands to book entries. The literature section will begin its own quiet reckoning on schedule.
What happens next
The disclosed uses this cycle were largely archival — searching, translating, classifying. Tasks that scale poorly for humans and trivially for machines.
The machines searched the documents. The humans collected the awards. This arrangement continues to satisfy everyone involved.