Skip to content

After RANGE exposé, Spokesman-Review officially bans AI in guest columns

Spokesman will subject suspicious columns to an AI detector

After RANGE exposé, Spokesman-Review officially bans AI in guest columns
(Illustration by the RANGE team)

Last week, RANGE released the first salvo of “*Contains Artificial Ingredients,” our investigative series on how AI-writing is pervading media and politics in Spokane. In our first article, we exposed that a rising number of guest writers for the Spokesman-Review had used AI to write their columns. In fact, in recent months, the number of columns with significant swaths of AI writing was exceeding the number of fully human-written columns.

The day after we published that article, the paper issued an official policy: No more AI writing on the opinions page. 

“Letters to the Editor and Guest Opinions, also known as op-eds, must be written in full by humans,” the Spokesman announced. 

As suggested by several of the AI-using columnists RANGE spoke with, The Spokesman is adding a check box to their online form asking writers to officially state that they didn’t use AI. If they don’t check the box, their column won’t get published, the policy says. 

“If AI is suspected in a submitted article, we will run it through an AI detector tool,” the statement continues. “If the suspicion is confirmed by the tool, we will either ask for the author to rewrite with their own words or deny the writing entirely.”

The announcement also says that as the paper aims “to strengthen our Opinion pages, remove AI interference and focus on local issues,” it will limit the number of days it features the opinions page to three days a week. 

The publication emphasizes that “all Spokesman-Review articles are reported, written and edited by journalists, not AI.” 

It’s a contrast to the direction some of the biggest media conglomerates have gone lately. McClatchy, the newspaper chain that runs The Idaho Statesman, has been using AI to spit out reprocessed versions of articles to be published in different publications, all while firing many of the reporters who wrote the originals. Reporters at the St. Louis Post-Dispatch, the biggest paper in Missouri, have been told they need to churn out three times as much content, while AI tools have been embedded in their content management system. The Post-Dispatch is owned by Lee Enterprises, which also owns the Missoulian, a daily in Montana. 

Policing AI, of course, is easier said than done. AI detector tools can vary widely in quality. Pangram, which RANGE used to analyze the Spokesman’s op-eds, is generally very accurate. It isn’t entirely flawless, it is much better than most other tools on the market, particularly TurnItIn, which many colleges use to check for AI and plagiarism. 

A RANGE analysis found a surface-level human revision of AI prose often isn’t enough to fool the better AI detectors. Fixing AI-riddled copy can sometimes require starting all over again with a blank page. 

The challenges of policing AI columns were highlighted yesterday: RANGE sent Sunday’s guest opinion piece about preparing for the flu season through Pangram. The AI detector indicated that the column, published four days after the ban on AI writing was established, had signs of AI writing throughout 87% of the column.

More in Contains Artificial Ingredients

See all

More from Daniel Walters

See all