"Parallel construction" is a technique where a journalist uses a private tip to guide an LLM's search across vast public data like podcasts and social media. The LLM then finds publicly citable evidence that confirms the private information, making the story reportable.

AI struggles to replace senior PR professionals because it lacks the nuanced, historical awareness to identify non-obvious risks. A human can spot a subtle connection, like a fallen soldier's link to royalty, that escalates a routine story into a major crisis—a connection AI would almost certainly miss.

AI can handle the 'writing lift,' much like historical rewrite desks. This forces a re-evaluation of a journalist's core value, shifting the emphasis from prose composition to the irreplaceable skills of investigation, sourcing, fact-gathering, and identifying what story matters.

Journalist Casey Newton uses AI tools not to write his columns, but to fact-check them after they're written. He finds that feeding his completed text into an LLM is a surprisingly effective way to catch factual errors, a significant improvement in model capability over the past year.

The foundation's own use of LLMs to analyze 3,000 disclosures showed that accuracy is highly sensitive to prompt design. Specificity, traceability, and continuous human oversight were essential to avoid misinterpreting varied corporate language and report structures.

California's CalMatters uses an AI called 'Tip Sheet' to analyze public records of politicians, including speeches, votes, and campaign contributions. The AI flags anomalies and potential stories, which it then provides exclusively to human journalists to investigate, creating a powerful human-AI partnership.

When pressed for sources on factual data, ChatGPT defaults to citing "general knowledge," providing misleading information with unearned confidence. This lack of verifiable sourcing makes it a liability for detail-oriented professions like journalism, requiring more time for correction than it saves in research.

The Atlantic's CEO Nick Thompson draws a clear line for AI in journalism. He advocates for using it extensively for reporting tasks like finding stories, analyzing data, or checking for chronological gaps. However, since a byline promises human authorship, AI should never write the final prose, even if it becomes a better writer.

A senior AI product manager at the Associated Press sparked controversy by suggesting reporters should focus on gathering quotes while LLMs handle the actual writing. This reflects a growing, contentious view among media leaders that devalues the craft of writing and reframes the journalist's role into data collection for an AI.