We have spent decades making journalism easier to publish, faster to distribute and simpler to consume. Yet we still ask an extraordinary amount of the people we serve: notice the story, recognise why it matters, remember what happened last time, find the update, and figure out what has actually changed.
We do the difficult work of reporting. Then we leave much of the work of staying informed to the reader. That is not just an audience-engagement problem. It is a limitation in how we have designed journalism itself.
An investigation can expose a failing institution. A report can reveal that a city has promised a new bus route, or that a school closure is being considered. But the question people live with rarely ends when the story does. Did the promise hold? Was the closure approved? Who benefited, who lost out, and what happens next?
A finished article is not always a finished public service. The most consequential possibility AI offers journalism is not that we can make more content. It is that we can finally begin to build journalism that stays with the questions people need answered.
The article is a form. Journalism is a function.
I often make this distinction in my work with newsrooms: verification, scrutiny, explanation and helping people navigate uncertainty are journalism’s functions. Articles, broadcasts, newsletters and podcasts are forms we use to deliver them.
Those forms matter. A narrative investigation can reveal a pattern no collection of personalised answers could capture. A shared story can create public attention and political pressure. The point is not to replace the article. It is to stop treating the article as the boundary of our responsibility.
Imagine a newsroom organised not only around stories to publish, but around public questions it commits to follow. Will the promised housing be built? What is happening to local water quality? How is a new policy actually affecting families? Each question has a history, sources, unresolved claims, competing accounts and future developments worth watching.
Reporting still creates the knowledge. But in a newsroom organised around questions, that knowledge becomes something more durable than a sequence of links: a living, accountable record that can be revisited, interrogated and updated as the world changes.
The invitation to the audience changes from “read this” to “we will keep watching this.” Not because every reader wants a personalized news feed, but because most people do not have the time to become expert navigators of every institution shaping their lives. This is a more demanding promise than an alert whenever a new article appears. It means noticing consequential change even when no new article was planned. It means sometimes saying that nothing has been resolved. And it means acknowledging when the newsroom can no longer reliably keep watch.
Making continuity easier to sustain
None of this began with AI. Community-centred newsrooms have long answered individual questions, followed public promises and built practical information services. To take one example: Outlier Media in Detroit has run a text-message service since 2016 that delivers housing and utilities information on demand and lets residents reach a reporter directly. It now uses AI for property lookups and to surface what residents are asking about, while journalists still write every message. It is journalism designed around what people need to know, not merely about what gets published.
What changes with AI is the range of connecting work that may become feasible. AI can help sift sprawling agendas, compare revisions to public documents, connect a new development to years of reporting, translate complex language, and adapt verified explanations for different levels of familiarity. It can help someone understand an issue without first knowing which agency, committee or legal term to search for.
The shift is subtle but profound as we move from publishing information and hoping people find their way through it, toward maintaining a service that helps them understand what matters as it unfolds. The newsroom still needs reporters who know which questions to ask, editors who judge significance, and clear records of where each claim came from. AI may lower the cost of connecting information but it does not remove the cost of being responsible for it.
In fact, the better these services become, the more important that responsibility becomes. If an error travels into a text message, translation or audio briefing, a correction cannot live only on the original article. This is where our editorial responsibility becomes product design.
The future of news is bigger than a chatbot
Much of the current conversation about AI and news begins with a familiar interface: a box where someone types a question and receives an answer. That can be useful. But it is an incomplete vision of journalism’s future.
A service built only around answering users’ questions leaves a crucial gap which is that people may not know what they need to ask. Journalism matters precisely because people often do not know what is happening, what information powerful institutions would rather they overlook, or what question they should be asking in the first place. An interface that answers perfectly formulated questions can still leave the biggest blind spots untouched.
This is why I am more interested in systems that combine responsiveness with editorial initiative. A person might want updates about a particular school. The newsroom can explain that school’s decision while also revealing how the policy affects other neighbourhoods, what evidence officials have omitted and why the wider choice deserves attention. Personal relevance should be a doorway into public understanding, not an escape from it. People may receive different formats, languages or levels of explanation, but not receive different underlying facts. Nor should an algorithm decide that an inconvenient truth is irrelevant because someone did not ask for it. And the service need not look like AI at all. It might arrive as a text message, a short audio explanation or a carefully timed update. Its form is irrelevant, as long as it makes journalism more useful.
What becomes of the newsroom?
There is a strategic question here as well as a civic one. If more people encounter reporting through assistants and platforms they do not associate with the publisher, newsrooms risk becoming invisible suppliers of raw material for somebody else’s product and business model.
Publishing more articles will not, by itself, solve that problem. A newsroom’s enduring value may lie increasingly in the knowledge it can stand behind: what it has verified, what remains disputed, what has changed, and which public questions it continues to investigate.
This also changes the work. The reporter remains indispensable, not as a human stamp on machine-produced prose but as the person who understands context, challenges official accounts, notices what is missing and earns trust. Editors become stewards of commitments that extend across time. Product and engineering teams help turn journalistic judgement into systems that can be tested, corrected and sustained. The newsroom becomes more than a place that produces coverage. It becomes an institution that remembers, checks and returns.
Sustaining this work still requires journalists who care whether a public promise was broken, earn trust in their communities and judge which changes matter. The most interesting use of AI is to make those human capabilities reach further.
The journalism we could not build until now
For much of journalism’s history, we have had to make difficult choices between depth and reach, between reporting a story and helping each person understand its implications, between following an issue and keeping thousands of people meaningfully informed as it develops. Technology does not eliminate those trade-offs. But it may change them enough to let us imagine services that once seemed impossible to maintain.
A newsroom that remembers every public commitment it has investigated. That connects a new decision to the reporting that preceded it. That can explain what changed in the language and format someone needs, without sacrificing a shared factual record. That notices the consequential development a reader never knew to search for. That knows when it is uncertain, when it has failed, and when it owes someone a correction.
This is not journalism made frictionless or automated into perfection. It is journalism whose value continues between publications. And perhaps the most important change is in how we understand success. It would mean asking more than how many stories we produced, how much attention they attracted or how efficiently we distributed them. Instead, the value of journalism would be measured by whether people were better able to understand the forces shaping their lives, recognise what had changed and exercise their own judgement.
The future worth building is not one in which AI does journalism for us, but one in which AI allows journalism to fulfil more of its purpose.