A website called Editor & Publisher has an article about how journalists are already using AI, so newsrooms need a policy to guide proper use. Skimming that gave me an idea…

What if part of journalism became a kind of record-keeping trail regarding the actual research that lead to a given piece? What if that even included detailed notes from AI prompts and results which contributed to a given piece? I recognize that much of what goes into developing a story might be confidential: talking to sources off the record, or following trails that lead to dead ends. Or having an editor axe certain parts of what you worked on – or sometimes an entire article. But what if that forensic trail could at least identify like [Source Protected] or [Redacted By Editorial] or for AI prompts, if they veered off topic into personal affairs could have sections marked [Redacted for Personal Data Protection].

Sure there would be technical questions to resolve, but there already are if we look into cases of for example questionable AI-assisted reporting like Margaux Blanchard. Right now, we have little to no transparency into cases like that, and this does nothing but further erode trust in the institutions caught in those webs. I’ve also seen in my own work, plenty of misreporting and completely false representation of “facts” which if we had more complete forensic tracking of the development of those stories (including for example archived versions of the pieces they were based on, and their metadata), it would be easy to identify where the errors were introduced.

Of course, I’ve rarely seen anyone address the underlying question and assumptions: does being a trusted brand in media actually improve your bottom line? Is “trust” essential to repairing what is essentially a system wide failure of media business models? I’m actually not so sure; it might be a puzzle piece, but it’s not going to stem the tide of the way things are headed… but is something like this still worth a try?

EDIT:

Here’s what I realized after writing this: people don’t even read the article, so why would they dive into some kind of forensic trail from which an article was composed in the first place?