AI for Journalists

Price: $49.99

journalist Journalist

Nobody at the municipality called it a cover-up. They didn't have to. The documents had been sitting in a filing cabinet for eleven months. The official who signed them retired in June. The one who should have flagged them was promoted in August.

Note

Note: This is forensic writing. The journalist is letting the chronology indict. No allegation is made directly.

ChatGPT ChatGPT

This is tight and deliberate. The structure mirrors the method: you are reconstructing a chain of non-action rather than asserting wrongdoing. The three-sentence sequence — documents sit, official retires, successor is promoted — functions as evidence, not editorializing. The opening line earns its irony because everything that follows justifies it. No changes needed to the logic. One small flag: "should have flagged them" implies a duty — confirm that duty is documented somewhere, or qualify it.

Claude — misreads the piece Claude

The opening line is risky. Starting with "nobody called it a cover-up" plants the word "cover-up" in the reader's mind even while denying it. This could expose you to accusations of bias. Consider a more neutral entry point that presents the facts without the rhetorical framing. The tone throughout feels prosecutorial — a more balanced approach would include a response from the municipality before drawing these inferences.

↓
ChatGPT responds to Claude's critique

ChatGPT Claude

Claude has misclassified this piece. It is not opinion writing that requires balance — it is forensic reconstruction. The journalist is not inferring a cover-up; she is documenting a sequence of inaction through verifiable dates and positions. The opening line is not a rhetorical trick; it is an accurate statement that reframes what follows. Calling this "prosecutorial" applies the wrong standard. The question is not whether the tone is neutral — it is whether the chain of events is accurately documented. That's a different test entirely.

↓
Claude responds to ChatGPT's critique

Claude ChatGPT

ChatGPT is right that I applied the wrong frame. I read this as advocacy writing and evaluated it for balance. That was a category error. Once I understand this as forensic reconstruction — where the method is to let the documented sequence speak — the opening line works precisely because it names what the evidence implies without asserting it. My concern about "prosecutorial tone" dissolves when the standard shifts from neutrality to evidentiary accuracy. The one flag that holds from my original critique: "should have flagged them" still needs a documented basis for that duty, or it is the one actual inference in an otherwise evidence-only paragraph.

↓
journalist adjudicates

Journalist Journalist

This is exactly what I needed to see. Claude's first response would have sent me toward a blander, more defensive version of the piece — and I might have accepted it without knowing why it felt wrong. ChatGPT named what I was doing before I had named it myself: forensic reconstruction. That framing is now something I can state explicitly when I brief any AI on my work. And Claude's revised point about "should have flagged them" is the one correction I'll actually make — the duty is in the municipal code and I can cite it. One sentence. Everything else stays.

Continue Reading

Enroll to unlock the full case study, resources, and all modules.