Journalism's Quislings for AI - And Why Writers Should Resist
"We live in capitalism. Its power seems inescapable… Any human power can be resisted and changed by human beings." That line from Ursula K. Le Guin stays useful because it names the point: resistance is a choice, especially in the art of words. Here's the full context.
If you ask people to pay for your writing, the bargain is simple: it's your thinking on the page. Outsourcing that to a pattern-matching machine breaks the bargain. Writing isn't a task to "optimize." Writing is the thinking.
Writing is thinking, not a step to outsource
Strip the writing out of reporting and you amputate the mind from the work. The draft is where you find the spine of a story, test your logic, and decide what matters.
Hand that draft to a large language model and you've ceded story shape, tone, and vocabulary to a system trained to average out voice. You can "fact-check" after the fact, but the decisions that define meaning are already baked in.
The cost-cutting mirage
Some editors now boast that AI "frees up an extra workday" by removing writing from reporters' workloads. That sounds efficient until you realize the savings come from gutting the core creative act.
Let's be plain: AI is an executive tool for cutting headcount and leverage. It doesn't ask for a raise, it doesn't go on leave, and it doesn't push back on interference. Calling its errors "hallucinations" is marketing sugar on top of lying machinery.
Readers can tell
Audiences aren't fools. They can spot the difference between a lived voice and generic slop. They don't pay for "content." They pay for perspective, judgment, and sentences with a pulse.
When a newsroom tells readers a model is doing the authoring, it's signaling contempt. When a byline hides that fact, it's worse.
Keep the craft where it matters
- Keep the first draft human. Notes, outlines, and drafts are where thought happens. Don't hand that over.
- If a team insists on AI drafts, refuse authorship credit. You're being asked to approve a product, not author a story.
- Disclose tools if management mandates them. Hidden automation erodes trust and your reputation travels with the byline.
- Protect writing time in scopes and contracts. Make "report, write, revise" explicit deliverables.
- Track error costs. Log corrections, time lost to fixing AI output, and the reputational hits. Bring data to negotiations.
- Set a style spine. Maintain a personal lexicon, banned-phrases list, and structural patterns that keep your voice intact.
If your newsroom pushes AI drafts
- Ask which model, which training data, and which guardrails. Vague answers are a red flag.
- Demand byline transparency. Readers deserve to know when machine-written text appears in the product.
- Insist on compensation for extra verification. If you're cleaning up model output, that's added labor.
- Propose a human-first policy: reporting, analysis, and all narrative decisions are writer-led; no machine-generated copy runs without clear labeling.
- Organize. This is a labor issue as much as a tech issue.
"But AI makes us faster" is the wrong metric
Speed without judgment is how errors ship. "More content" is not a strategy; it's an inventory problem disguised as innovation.
If you need a benchmark, look to standards bodies and how they frame disclosure, attribution, and limits. The Associated Press guidance is a decent reference point for newsrooms figuring this out: AP on generative AI in news.
It's not "ultra-processed." It's poison to discourse
Machine-made language floods the feed with confident nonsense. That pushes the conversation toward noise and away from accountability. At scale, it sickens civic life.
We already tolerate trace amounts of error in human work. Choosing a diet of nothing but machine slurry is a different proposition. Don't help normalize it.
Practical habits to strengthen your edge
- Do zero-assist writing sprints. 45-90 minutes, no tools, no tabs. Build the muscle that gives you an advantage.
- Annotate your reporting trail. Show your receipts. It sharpens your argument and speeds edits.
- Read outside your beat daily. Better inputs, better sentences.
- Publish under your real voice. Consistency compounds trust and makes you hard to swap for a template.
Know the tech you're rejecting (or restricting)
Understanding how large language models work helps you argue for sane limits and spot canned sales pitches. If you want a primer on how these systems generate text and where they fail, see this overview: Generative AI and LLM.
The line worth drawing
Tools can help with transcripts, search, or formatting. But authorship is not a clerical task. The draft is where your judgment lives.
Hold that line. Your readers will feel it. Your career will thank you.
Your membership also unlocks:
