Nieman Lab.
Predictions for
Journalism, 2025.
For younger colleagues, auto-generated feeds may be all that internet culture has ever offered — especially on apps that order posts based on engagement. But there was a time when social media was mostly all in reverse chronological order. It was…unpleasant, but it was free.
Feed algorithms were introduced to address what sounds like a quantum physics problem: What’s the difference between when an event occurs and its relevance? The approach transformed how the news business worked, and today almost everything online is powered by some version of user experience, data collection, analysis, and recommendations.
For a while, things felt more useful this way — more curated, less awkward. The experiment had worked, we thought, as news stories took the top spots in every social feed. Publishers who posted on social media scrambled to build relationships with platforms that seemingly respected the role of journalism in democracy. Many conferences were held. Did audiences understand how or why particular stories were served specifically to them? Unclear.
During this bucolic era, the role of “editorial judgment” quietly faded.
Curation seemed doomed by the time publishers focused more on what was “most read” than what editors felt was “most relevant,” and even today the two categories haven’t made peace. Most early social media editors, like me, hailed a mixed approach as a celebration of audience relatability and as a new avenue to build trust between newsrooms and the public, which was already eroding by the mid-2000s.
We asked: If audiences knew that newsrooms respected what they read (or watched, or otherwise engaged with) the most, would it create some sense of community? It was a guess and, for a time, it worked so well that billion-dollar startups were born out of boozy media parties. With new metrics created to align “old” and “new” media revenue models, the business of social media engagement optimization boomed. It eventually turned into what we call audience development, growth, and optimization.
Along the way, tens of thousands of journalists with hundreds of thousands of years of combined editorial sense and experience were laid off. That wasn’t part of the deal, was it? Eventually, platforms moved on from prioritizing news, and publishers were largely cast aside for entertainment media.
As it turned out, prioritizing news stories wasn’t always part of the deal with platforms. Many publishers have gone bankrupt since platforms stopped prioritizing news. Since 2016, it’s been unclear what the future of digital publishing would be without paid digital marketing — this was the transition from organic to paid you may hear if you speak to a marketer. 2015 will be a full decade since the tsunami of news audiences first slowed down. Engagement has mostly struggled to find the same reach since feeds became more “useful.”
Many of the jobs that were created to optimize for algorithms were eventually scrapped, too. But something else unexpected happened: Large language models enabled tech companies to imagine media distribution with even fewer people involved. AI will soon replace older feed algorithms and threatens to further weaken the role of editorial judgment in online publishing.
It won’t be long before AI is used to generate new metrics and new analysis, and to push user behavior further away from curated editorial experiences. I’m sure of this, because that’s exactly what social algorithms did. (For the record, there are many reasons to be excited about AI, but — as someone who values the creative aspects of media to build audience — I’ll focus on curation alone here.)
What stops us from expanding the last innovative module of the homepage, “Most Read/Viewed”? Publishers will soon access AI-created analysis that will “predict” what is most likely to be read. Who will be the first editor to assign stories using not judgment or experience but prediction — based on a database they’ll never fact-check themselves? Doing so will have an impact orders of magnitude more disruptive than anything in the past 20 years. Journalists are tasked to explore the new and unexamined — not gamble with our trade’s ethics by using data we don’t verify.
Future journalists must answer these questions:
How does creativity help us discover new stories, or rethink old ones?
Can we preserve the artistic qualities of media while also pursuing the best options for the next era of the digital media business?
What are the ethics of assigning stories using predictive data?
Meanwhile, continue sending friends links to stories you wrote or produced. Share with colleagues what media you like, and why. Develop taste in writing and formats. Buy that print copy. We may be in the last decades when we do without a chatbot suggesting the right way to curate. Today, I asked Gemini if I should use AI to write stories, and it suggested I consider the limitations of AI before noting it’s a “personal choice.” Every chatbot needs an editor, I say.
Margarita Noriega is a managing editor at Morning Brew.
For younger colleagues, auto-generated feeds may be all that internet culture has ever offered — especially on apps that order posts based on engagement. But there was a time when social media was mostly all in reverse chronological order. It was…unpleasant, but it was free.
Feed algorithms were introduced to address what sounds like a quantum physics problem: What’s the difference between when an event occurs and its relevance? The approach transformed how the news business worked, and today almost everything online is powered by some version of user experience, data collection, analysis, and recommendations.
For a while, things felt more useful this way — more curated, less awkward. The experiment had worked, we thought, as news stories took the top spots in every social feed. Publishers who posted on social media scrambled to build relationships with platforms that seemingly respected the role of journalism in democracy. Many conferences were held. Did audiences understand how or why particular stories were served specifically to them? Unclear.
During this bucolic era, the role of “editorial judgment” quietly faded.
Curation seemed doomed by the time publishers focused more on what was “most read” than what editors felt was “most relevant,” and even today the two categories haven’t made peace. Most early social media editors, like me, hailed a mixed approach as a celebration of audience relatability and as a new avenue to build trust between newsrooms and the public, which was already eroding by the mid-2000s.
We asked: If audiences knew that newsrooms respected what they read (or watched, or otherwise engaged with) the most, would it create some sense of community? It was a guess and, for a time, it worked so well that billion-dollar startups were born out of boozy media parties. With new metrics created to align “old” and “new” media revenue models, the business of social media engagement optimization boomed. It eventually turned into what we call audience development, growth, and optimization.
Along the way, tens of thousands of journalists with hundreds of thousands of years of combined editorial sense and experience were laid off. That wasn’t part of the deal, was it? Eventually, platforms moved on from prioritizing news, and publishers were largely cast aside for entertainment media.
As it turned out, prioritizing news stories wasn’t always part of the deal with platforms. Many publishers have gone bankrupt since platforms stopped prioritizing news. Since 2016, it’s been unclear what the future of digital publishing would be without paid digital marketing — this was the transition from organic to paid you may hear if you speak to a marketer. 2015 will be a full decade since the tsunami of news audiences first slowed down. Engagement has mostly struggled to find the same reach since feeds became more “useful.”
Many of the jobs that were created to optimize for algorithms were eventually scrapped, too. But something else unexpected happened: Large language models enabled tech companies to imagine media distribution with even fewer people involved. AI will soon replace older feed algorithms and threatens to further weaken the role of editorial judgment in online publishing.
It won’t be long before AI is used to generate new metrics and new analysis, and to push user behavior further away from curated editorial experiences. I’m sure of this, because that’s exactly what social algorithms did. (For the record, there are many reasons to be excited about AI, but — as someone who values the creative aspects of media to build audience — I’ll focus on curation alone here.)
What stops us from expanding the last innovative module of the homepage, “Most Read/Viewed”? Publishers will soon access AI-created analysis that will “predict” what is most likely to be read. Who will be the first editor to assign stories using not judgment or experience but prediction — based on a database they’ll never fact-check themselves? Doing so will have an impact orders of magnitude more disruptive than anything in the past 20 years. Journalists are tasked to explore the new and unexamined — not gamble with our trade’s ethics by using data we don’t verify.
Future journalists must answer these questions:
How does creativity help us discover new stories, or rethink old ones?
Can we preserve the artistic qualities of media while also pursuing the best options for the next era of the digital media business?
What are the ethics of assigning stories using predictive data?
Meanwhile, continue sending friends links to stories you wrote or produced. Share with colleagues what media you like, and why. Develop taste in writing and formats. Buy that print copy. We may be in the last decades when we do without a chatbot suggesting the right way to curate. Today, I asked Gemini if I should use AI to write stories, and it suggested I consider the limitations of AI before noting it’s a “personal choice.” Every chatbot needs an editor, I say.
Margarita Noriega is a managing editor at Morning Brew.
