CNET used AI to write articles for a year and a half – now they're pumping the brakes
For no less than 18 months, CNET has employed AI tools to give financial advice, but outside reporting has made it pause
➡️ The Shortcut Skinny: Do no harm
🤖 CNET’s use of AI to produce financial advice started at least 18 months ago
🤔 Few inside the company knew about the AI tool being used
🛑 CNET is pausing the program for now
✉️ Associated Press has used automated writing since 2014
🏈 AP and others regularly report on low-level sports games using automation
💰 CNET’s use, though, doles out flawed, potentially harmful financial advice
Last week, CNET came clean about its use of AI to craft search engine-optimized personal finance stories, the ongoing analysis of which has revealed glaring, basic math errors that should’ve been caught by the human editors the site says are reviewing the articles for accuracy. Today, we found out the site’s been running its automated writing “experiment” for at least a year and a half – far longer than the roughly two months we initially thought.
Reporting this morning exposed an internal e-mail from former CNET editor Rae Hodge, who wrote on her last day encouraging her co-workers to be openly critical of CNET parent company Red Ventures’ use of AI for published content without attributing it as such. The Verge acquired the e-mail, along with anonymous statements from current and former staff members, who say they were asked to use the tool to churn out SEO-friendly content.
Hodge reportedly also went on to say the AI tool was being used to craft content for cybersecurity newsletters, and were filled with potentially harmful errors.
Few inside CNET were keenly aware of the tools outside of a small CNET Money team led by Managing Editor Justin Jaffe, who has been with the outlet since 2015. Writers outside of the team reportedly had difficulty telling when an article was AI-generated or written by a human colleague.
Don’t call them reporters
Following its reporting, The Verge published a subsequent story (just as we were prepping this article to go live, in fact), detailing a staff call that took place today announcing that CNET would be pausing its AI writer program for now.
On the call, however, editor-in-chief Connie Guglielmo told staff, “We didn’t do it in secret; we did it quietly.” Articles written by what, it turns out, was an AI tool created by Red Ventures for this purpose, were not disclosed as machine-created until after Futurism broke the story. You know, like a secret.
The Verge article then quoted Guglielmo as saying, "Some writers – I won’t call them reporters – have conflated these two things and had caused confusion and have somehow said that using a tool to insert numbers into interest rate or stock price stories is somehow part of some, I don’t know, devious enterprise.”
She then pointed out several publications known to use AI to add data to financial reporting. Notably, prior to the discovery, hovering your mouse over the articles’ ‘CNET Money Staff’ byline would display the text: “This article was generated using automation technology and thoroughly edited and fact-checked by an editor on our editorial staff,” implying more than simple data injection.
Automation in journalism isn’t new
CNET’s moves have been part of a growing trend in the use of AI text generators. The Verge added in its report that The Associated Press started using the same Wordsmith tool as CNET nearly a decade ago – and more recently, one called Data Skrive – using vague terminology to denote where its articles used AI to report on corporate earnings and sports statistics.
AP automated sports articles like one detailing a then-upcoming sports match in November 2022, lack any sort of crucial context to help the content be useful. What we can glean from the story is that the Utah Tech Trailblazers – which is presumably a college (but you wouldn’t know it from the article) – would soon play the Idaho Vandals in a game. What game? I guess the text generator didn’t think that was relevant.
The generator gave up some details about the teams’ previous season, but confusing terminology and an unexplained acronym make it hard to fully know what it’s trying to say.
At the bottom, AP attributes the article to technology from Data Skrive, an SEO content creation automation company specializing in sports, with over 250 article types that it boasts have been used by ESPN, USA Today, Sports Illustrated and more.
CNET’s money farms
Like CNET, The Associated Press pitches its use of automation as a net benefit for readers and its workers, freeing up journalists to work on more in-depth reporting.
The difference, however, is that while AP has apparently been using its tools for basic data reporting, CNET’s application has seemingly been more cynical: creating advice articles with the specific goal of farming money from its readership. Although the outlet says the articles are reviewed by human editors for errors, they don’t seem to be doing a great job, resulting in misleading claims that could spell disappointment and disaster for its readership.
All of the articles credited to CNET Money – which are the ones written using AI – now feature a disclaimer at the top, stating, “We are currently reviewing this story for accuracy. If we find errors, we will update and issue corrections.”