A record number of Pulitzer awardees disclosed AI use this year, marking a clearer view of how major news organizations are bringing large language models into prize-level work. According to Nieman Lab, eight awardees made disclosures: five winners and three finalists.
The disclosures do not point to a single standard workflow. Instead, they show a narrower pattern: AI was used mostly to move through large collections of material faster, while the final journalistic judgment remained tied to human review, classification, verification or expert checking.
What The Pulitzer Disclosures Show
The Pulitzer Prizes have required AI disclosures since 2024. This year, entrants used AI tools and large language models more often, mainly to search large document sets faster.
That distinction matters. The reported uses were not presented as AI replacing the reporting process or writing prize submissions. They were described as tools for handling scale: thousands of public documents, tens of thousands of leaked documents, a difficult diary translation and a check against a manual classification process.
The record number also shows that disclosure is becoming part of the public record around major journalism awards. A disclosure requirement does not decide whether a use is good or bad by itself. It creates a way to see where AI entered the work and how newsrooms say they controlled it.
How Newsrooms Used AI
The examples reported by Nieman Lab show several distinct uses of AI in Pulitzer-recognized work.
- The Wall Street Journal used an internal LLM to summarize thousands of public documents on Texas floods.
- The Minnesota Star Tribune used ChatGPT to translate a female shooter's diary written in faux Cyrillic, then had language experts check the results.
- The Associated Press used an LLM to search tens of thousands of leaked documents on Chinese surveillance technology.
- The New York Times used GPT-5 to check its manual classification of SEC crypto cases.
Those uses are different, but they share a common feature: each begins with a large or difficult source base. Public records, leaked documents, unusual writing systems and classified case sets can all create time-consuming review work. In the examples disclosed, AI helped with finding, summarizing, translating or checking information inside those materials.
The Minnesota Star Tribune example is especially clear about the role of human verification. ChatGPT was used for translation, but language experts checked the results. That detail limits what can reasonably be inferred: the tool helped produce a translation, but the process still included expert review.
The New York Times example also shows AI being used as a check against work already done by people. The newsroom used GPT-5 to check its manual classification of SEC crypto cases. Based on the source, that points to AI as a review aid rather than the sole classifier.
Why Document Work Is The Main Pattern
The most visible theme is document-heavy journalism. The Wall Street Journal summarized thousands of public documents on Texas floods. The Associated Press searched tens of thousands of leaked documents on Chinese surveillance technology. These are tasks where the amount of material can become a major practical obstacle.
Large language models are useful in these examples because they can help reporters move across text at speed. The source says entrants used AI tools and large language models more often, mainly to search large document sets faster. That is a limited but important claim: the emphasis is on navigation through information, not on handing over editorial control.
Summarizing documents, searching leaked files and checking classifications are not identical tasks. A summary compresses material. A search helps locate relevant material. A classification check compares a categorization process against another pass. Each use carries different risks, which is why disclosure and review matter.
The examples also show why AI use in journalism is not one conversation. A tool used to search documents raises different questions from a tool used to translate a diary. A tool used to check manual classification raises different questions from a tool used to summarize public records. The disclosed Pulitzer cases make those differences easier to discuss because the uses are attached to concrete newsroom workflows.
Where Pulitzer Officials Draw A Line
Pulitzer administrator Marjorie Miller told Nieman Lab that the news industry now accepts "AI is here to stay" and better understands where its use is appropriate and "when it might not, such as in writing and editing stories in any format that might be considered for a Pulitzer Prize."
That statement captures the balance now emerging around AI disclosures. The issue is not simply whether a newsroom used AI. The more important questions are what the tool did, where humans checked the work and whether the use affected writing or editing in a way that would raise concerns for Pulitzer consideration.
The disclosure rule is also expanding. According to the source, it will cover book entries starting next year. That means the same transparency requirement applied to Pulitzer journalism entries since 2024 will move into another part of the prize process.
For readers, the practical takeaway is straightforward: AI is already present in some award-recognized work, but the disclosed examples show a preference for behind-the-scenes assistance with large amounts of source material. The record eight disclosures make that presence more visible, and the Pulitzer requirement gives the public a clearer basis for judging how these tools are being used.