AI Tries To Write The News: What Writers Can Learn This Week (Dec. 7)
This weekly experiment pits AI-written drafts against real local events. The point isn't to fact-check the bot - it's to show writers where AI slips and how to sharpen our work.
Two stories stood out: Alaska's proposed state budget and a new Emmy nod for Vera Starbard. Both reveal the same lesson - specificity wins; filler loses.
Case 1: Alaska's budget proposal - the numbers are the story
The real update: a proposed budget projecting a $1.5B-$1.7B deficit, built on a "statutory" PFD and a 10-year plan promising ~$15B in new revenue with details to be revealed next month. It also broke tradition with no press conference - just a short Facebook post.
The AI draft glossed over the hard parts. It swapped numbers for generic lines about "education, healthcare, infrastructure, public safety," then nodded at "fiscal discipline" without proof. It read like a brochure, not a budget brief.
- Lead with the numbers: $1.5B deficit, $3,650 PFD, $15B "new revenue." That's the hook.
- Interrogate the premise: If revenue is "to be named later," say so - clearly.
- Spot what's unusual: No press conference for a final-term budget is newsworthy.
- Link to source docs: If available, point readers to official materials like the Alaska OMB.
- Kill boilerplate: Replace vague "funding for education" with exact line items or the absence of them.
Case 2: Vera Starbard's Emmy nod - details build trust
The real update: State writer laureate Vera Starbard earned her fourth nomination for Outstanding Writing for a Preschool Animated Series on PBS's "Molly of Denali." The episode, "Thanks-for-giving," explores Indigenous perspectives on the holiday and its history. The series features the first nationally distributed children's show with an Alaska Native lead and has a Peabody and a 2025 Children's & Family Emmy to its name.
The AI version flattened all that into generic praise, then paired it with a wildly off-base "photo." It skipped the category, the episode title, and the show's milestones - the pieces that make the story matter.
- Name the category and episode: Don't bury the exact win path.
- Add the timeline: Past awards and why this nomination lands now.
- Avoid invented visuals: If you can't verify an image, don't run it.
- Anchor with an official link: The show's page on PBS KIDS keeps readers grounded.
Quick tells an article was drafted by AI
- Hedged generalities instead of hard numbers.
- Invented "highlights" that weren't in the source material.
- Safe adjectives, zero attribution.
- No links to primary sources or official documents.
- Reads like a template: same four sectors, same tone, every time.
A pre-publish checklist for writers
- Numbers: Exact, sourced, and consistent throughout.
- Sources: Named people, documents, or pages readers can visit.
- Nouns over adjectives: Categories, dates, amounts, titles.
- Quotes: Verifiable and contextually accurate.
- One-line summary test: If you can't state the core claim in a sentence, the piece isn't tight enough.
Use AI - without letting it write your story
AI is useful for structure and speed. It's weak on truth and nuance. Keep it on a leash.
- "List the 10 most contentious facts or numbers in this story - no filler."
- "Give me 8 skeptic questions a reader would ask about this claim."
- "Summarize this budget in one sentence for a busy parent."
- "Offer 5 angles that put people, not policy, at the center."
If you want deeper reps on prompts and workflows for writers, these guides help: Prompt engineering for writers and AI courses by job.
Bottom line
Specifics separate reporting from fluff. Lead with facts, trim the filler, and link to something real. Let AI assist - don't let it decide what matters.
Your membership also unlocks:
