AI is everywhere these days, but when you see an article online from what has historically been a trusted source, it’s reasonable to think an AI wasn’t involved. At major newsrooms across the country, though, that reality is increasingly becoming less clear cut.
While not every use of AI in the newsroom is as blatant as an AI just drafting a post onto a blank page (although that happens too,) it’s important to know just how the sausage is being made when it comes to the information that’s shaping your worldviews.
Here are just a few news outlets that have started openly using AI in their processes, so you can stay informed about where exactly the information you’re reading is coming from.
The New York Times
Earlier this week, Semafor published an article about new AI tools that New York Times management is reportedly encouraging staffers to try. While the publication’s legal arm is currently embroiled in a lawsuit against OpenAI and Microsoft for copyright infringement, it appears the site’s editorial and product wings are ready to embrace AI in the newsroom, at least going by communication from the top.
According to Semafor’s writeup, NYT management is now supplying AI training to its journalists, debuting an internal AI tool called “Echo,” and approving usage of external AI tools including Google’s Vertex AI, a few Amazon AI products, and ironically, Microsoft’s Copilot and a non-ChatGPT tool from OpenAI.
Not all of these will be used for the website’s articles—the New York Times is bigger than what you see on its front page—but Semafor did say that journalists are being encouraged to “use these AI tools” for tasks like mild content revisions or coming up with questions to ask during interviews.
“Generative AI can assist our journalists in uncovering the truth and helping more people understand the world,” read the company’s editorial guidelines on AI, which are public on its site. “We view the technology not as some magical solution but as a powerful tool.”
To that end, there are seemingly still some guardrails still in place. Semafor says the NYT has warned staff not to use AI to draft or significantly revise articles, and has noted that AI use could potentially infringe on copyright or unintentionally expose sources.
Still, with internal communication suggesting writers use AI to come up with headlines and draft social copy, it’s worth keeping hallucination in mind the next time you see a suspicious-seeming NYT story going viral on social media.
For the journalists’ part, Semafor reports that some remain skeptical, worrying that AI in the newsroom could inspire “laziness” and reduce accuracy or creativity.
Quartz
First off, a disclaimer. Quartz is currently owned by G/O Media, which also owned Lifehacker prior to its sale to Ziff Davis in 2023.
With that said, G/O Media has since then become a major proponent of AI in the newsroom, with business news outlet Quartz being its biggest experiment in this.
Scroll through Quartz’ bylines for a bit and you’ll find posts attributed to the Quartz Intelligence Newsroom, which seems to be dropping any of the restraints adopted by the New York Times. Here, the AI “writer” has been quietly generating earning reports for months and has recently started spitting out more general blogs as well. These include stories about potential Bitcoin value or how to delete your Meta-owned social media accounts, but as you might expect, there are issues to be aware of with each.
To Quartz’ credit, it does not hide that these stories are AI-generated and the AI cites its sources, but it doesn’t appear as if there’s human oversight addressing any problems that might arise from this.
For instance, the Quartz article on how to delete your Meta-owned social media accounts seems to be a simple regurgitation of a TechCrunch story (which did not give permission to be summarized, for what it’s worth), but with clear instructions swapped for what the cribbed TechCrunch writer calls “vague” gestures in the right direction. Speaking about other stories written by the Quartz AI, the same writer also said “my editor would never let me publish something so sloppy.”
With that, it’s very likely you might find this story hoping for useful advice, only to leave disappointed and more confused than when you came. But the chances of that only get higher if you land on another story with a less reliable source than TechCrunch.
As noted by Futurism, the Intelligence Newsroom has frequently cited a site called Devdiscourse, which itself has all the appearances of an AI content farm. When robots are citing robots, I do have to ask: why not just go to ChatGPT and prompt it yourself? Even if I accept AI news as worthwhile, it’s unclear to me what G/O’s efforts are adding here.
G/O is slapping a disclaimer on all AI-generated Quartz stories that says it’s in the “first phase of an experimental new version of reporting,” but with AI efforts at the company going back to 2023, it’s unclear if it’ll ever figure out what that experiment is supposed to yield.
In the meantime, keep an eye out on all Quartz bylines, and if you sniff AI, maybe consider giving the sources the bot is blending up for you a read instead. It seems like once you click that attractive headline, you’re basically playing a game of roulette.
AP
If the New York Times is just starting to dip its toes into AI, and Quartz has done a full-on cannonball, then the AP’s use of AI seems to be somewhere in between. On its site, the agency proudly declares that it uses AI for translation, transcription, headlines, research, and even some automated articles, but general blogs are still left to human hands.
“Our goal is to give people a good way to understand how we can do a little experimentation but also be safe,” said AP’s Vice President of News Standards and Inclusion, Amanda Barret, in 2023, when the group first issued its guidelines on artificial intelligence.
Where this will most likely affect you is in the site’s use of Wordsmith, an AI program that specializes in summarizing content like sports scores, weather reports, and, as with Quartz, earnings. AP has been using a version of this program since 2014, so it’s not exactly new, but it’s worth being aware that stories on these topics without a specific author attached likely came from a bot. But aside from these stories, AP’s only other clear use of directly AI-written content was an experiment with reporting public safety incidents in a specific Minnesota newspaper.
Otherwise, the only time you’re likely to see direct AI content on an AP story is its story summarizing pilot: for instance, AI might put a blurb under election day stories to say which offices are up for grabs.
Again, headlines and research are still admittedly AI-assisted, so it’s important to be extra careful when something doesn’t pass your sniff test, but it seems like humans are still taking the forefront here, at least for now.
The Washington Post
The Washington Post’s use of AI is unique, in that it doesn’t directly impact the content. Rather, it’s more of an enhanced search engine for readers who go looking for it. The bot, called “Ask the Post AI,” takes questions, spits out a brief AI-generated response trained on Washington Post content, and then lists relevant articles below in a Google-like manner.
“Answers are AI-generated from published reporting,” warns the bot when you ask a question. “Please verify by consulting the provided articles.”
Responses seem to aim for a measured tone, leaning on quotes from noteworthy sources in matters of opinion, and when I asked the bot about President Trump’s evolving stance on TikTok or Elon Musk’s history with Tesla, I got truthful responses back, although the former was a paragraph long while the latter was only a sentence long. I suppose how detailed your responses are depends on how much the paper has covered a specific topic.
Washington Post hasn’t been clear about which AI tech powers Ask the Post AI, but in general, it won’t bother you unless you go looking for it. Based on the prevalent warnings about its use, it seems to be intended more as a place to start research than as a direct news replacement.