Contrary to popular belief, most investigative journalism, including landmark cases like Watergate, is not the result of reporters digging for stories. Instead, it originates from sources who approach journalists with a specific story they want publicized, which inherently skews the entire news ecosystem.
Many media outlets misapply AI by trying to make it write articles, which results in terrible quality. The technology's true journalistic value lies in its unprecedented ability to analyze vast troves of legal and financial documents, uncovering stories that were previously impossible for humans to find.
Journalists should report on investigations that yield no findings. This practice, while unnatural for humans, serves as a crucial check on motivated reasoning and conspiracy theories. It provides evidence of absence after a thorough search, which is a vital but often overlooked form of knowledge.
AI can defeat financial obfuscation techniques, like using Swiss foundations as intermediaries. Even if a recipient's funding is hidden, AI can scan the public disclosures of potential donors, find matching transaction records, and successfully trace the money back to its source.
Contrary to traditional journalism where narrow hypotheses are key, AI-driven investigations yield better results from broad, open-ended prompts. This approach inverts the old workflow, allowing the AI to surface unexpected connections from the data rather than merely confirming a reporter's preconceived idea.
While AI-generated stories could be used to cherry-pick facts, this is an improvement over the current media ecosystem where sources and facts are often entirely fabricated. Grounding stories in verifiable documents imposes a discipline and falsifiability that is largely absent today.
Large Language Models are often more persuasive than humans. Research suggests this is not due to manipulation, but because they can embody classical rhetorical virtues perfectly: being infinitely patient, non-condescending, and empathetic—traits humans struggle to maintain consistently in debates.
The real 'apocalypse' was the long history of unchecked destructive behavior and flawed arguments. AI's most significant societal impact will be to act as a historical corrective, systematically revealing past institutional failures and lazy logic, rather than creating a new crisis.
The post-Nixon push for government transparency didn't increase accountability as promised. While more data became available, news organizations, still reliant on source-driven reporting, never developed the infrastructure to effectively analyze it. AI is the first technology to finally bridge this gap.
The economic disruption of AI in journalism isn't about incremental cost savings. Its real power is enabling stories that were previously 'straightforwardly infeasible' to produce, regardless of budget. These are investigations that would have required an impossibly large team of specialists, costing tens of millions.
Today's societal problems aren't a 'crisis of liberalism,' but a 'crisis of nominalism.' We take labels like 'science,' 'accountability,' or 'auditing' for granted without examining the underlying reality. Many institutions have become hollow shells, and we've failed to look 'inside the box.'
A significant threat in the AI era is not just disinformation, but mass data obfuscation. Adversaries will 'maliciously comply' with transparency laws by burying truth in petabytes of AI-generated 'slop.' This creates a 'Red Queen's race' where truth-seekers must constantly innovate to keep up.