I hoped 2025 would bring clarity about how generative artificial intelligence would change the media. It didn’t.
There was no “deepfake apocalypse,” only the steady flow of AI slop onto our social media feeds. AI companies signed licensing deals with news organizations, but lawsuits between the two parties remain unresolved. And newsrooms continued their cycle of loud failures and quiet experimentation with the technology.
Two stories this year did come pretty close to helping me understand AI’s impact. One was The Washington Post’s relentless AI rollout of AI-driven products. The other was the Walt Disney Company’s threats against — and eventual $1 billion agreement with — OpenAI to allow Mickey Mouse, Luke Skywalker and Simba to be AI-slopified in Sora.
In those stories, and across the peaks and troughs of AI in journalism this year, I found one throughline: guardrails. Who built them, who ignored them and what happened when they failed. “Move fast and break things” versus “seek truth and minimize harm” has always been an easy choice in journalism. But as we compete with AI chatbots that answer questions instantly and search engines that bury our links, that choice is getting harder.
To survive, newsrooms have to experiment, take risks and push boundaries — but do it without breaking things that separate us from the slop: humanity and trust.
So, as we close out 2025, I offer my cheers to those who found that balance — and jeers to those who didn’t.
Cheers: The Minnesota Star-Tribune, for its smart use of AI to decode and report the videos and journal pages published by the Annunciation Church shooter. I use this example in every AI training, because it shows that the technology — used thoughtfully and with human oversight — can allow journalists to do quick and important investigative work.
Jeers: The Chicago Sun-Times, for its summer reading guide, which contained made-up book titles written by ChatGPT. Remember folks, freelancers should sign on to your AI ethics policies.
Cheers: The Onion and CEO Ben Collins, who quickly apologized for an undisclosed AI image used in one of the satirical publication’s articles in early 2025. Collins has taken a stand against any AI use at the Onion, which I respect. AI guidelines should align with an organization’s values. And AI can’t be funny.
Jeers: Business Insider and CEO Barbara Peng, for declaring it was going “all-in on AI” while also announcing major layoffs. At a time when journalists are justifiably worried about large language models stealing their content and replacing them, it was a tone deaf move that made my job teaching AI literacy in newsrooms a lot harder.
Cheers: The Washington Post, for relentless experimentation with AI tools. I know I’m going to get absolutely flamed for this one (keep reading before you start typing). But I think the Post has recognized the huge value in its archives and experimented with ways to squeeze more out of it: Climate Answers, its AI search and comment summaries and its personalized podcast.
Jeers: The Washington Post, for the haphazard rollout of most of the same products. I remain a big fan of former AI editor Phoebe Connelly’s approach to Climate Answers, which included a feedback form. I have yet to find something similar on any other Post AI products, including its apparently error-prone personalized podcasts. The Post should probably update its AI policy, which doesn’t outline these types of uses. It currently states: “This means using AI to suggest related content, to translate between languages, or to sift through pages of documents or thousands of images. Whatever the task, we recognize this is a new and imperfect technology — meaning accuracy cannot be assumed and verification is critical.”
Cheers: Wikipedia, and its volunteer editors who have faced the daunting task of setting AI policies and rooting out AI slop on the online encyclopedia. I use its guide on identifying AI writing often, and regularly scope out its reports on AI activity on the site.
Jeers: Grokipedia. Enough said.
Cheers: The New York Times, for building AI tools like Echo and Cheatsheet that help reporters do stellar digital investigations — like analyzing President Donald Trump’s cognitive health via troves of documents, or breaking down the Sydney Sweeney jeans ad controversy.
Jeers: “Margaux Blanchard,” and the publications that published AI-generated articles from the nonexistent freelancer. AI tools like ChatGPT and Claude are only getting better at “writing,” so editors have to be more diligent about fact-checking.
Cheers: Google, for finally integrating its SynthID into Gemini. The digital watermark means it’s easier to figure out if something has been created with Google’s incredibly powerful Nano Banana image generator.
Jeers: Meta and OpenAI, for creating digital troughs of AI slop in their feeds of AI-generated videos. The videos are disturbing, and dangerous when they break containment. Sora and Meta AI were rolled out with little regard for user safety or the information ecosystem, with hilarious but mostly heinous results.
Cheers: The Associated Press, for hosting dozens of news leaders at the second Poynter Summit on AI, Ethics and Journalism in April. In an industry where everyone’s experimenting in silos, AP helped create a space to develop collective guardrails.
Jeers: Fox News, for falling for AI-generated videos of SNAP beneficiaries threatening to “ransack stores” during a government shutdown. Fox ran the story, then quietly rewrote it with an editor’s note admitting “some videos that appear to have been generated by AI” after commentators called them out. Rage bait is the Oxford Word of the Year, and it’ll only get worse as AI-generated videos get more convincing.
Cheers: The Philadelphia Inquirer, for Dewey, its open-source AI archive research assistant. It inspired me to build something similar for PolitiFact, and should hopefully do the same for every newsroom looking to get more out of its valuable archives.
Jeers: The Los Angeles Times, and its AI experimentation in its opinion pages. In less than a week, the publication launched the tool, watched it defend the KK and then shut it down.
Cheers: U.K. fact-checker Full Fact, Fundación Maldita.es and the European Fact-Checking Standards Network Prebunking at Scale Project, which aims to catch misinformation before it goes viral. As someone who spends a lot of time prebunking — via my media literacy work with MediaWise — this is a noteworthy use of AI in journalism.
Jeers: Grok. Enough said.
Cheers: 404 Media, and its stellar coverage of all things AI. You’ll find their links throughout this piece.
And, finally…
Jeers: Me, for starting a list that can’t possibly capture the brilliant work being done by journalists with the help of AI, the support organizations helping out, the people I’ve met out and about and the funders paying for a lot of it. That’s a good problem to have. Thanks for a great 2025, y’all.
