For two decades, digital news has been trained to chase the click. In the Attention Economy, the rule was simple: keep people looking and sell that attention. That logic pushed editorial strategies toward volume, speed, and SEO. It wasn’t a moral failure so much as an incentive trap. The system rewarded raw screen time, so the industry optimised accordingly.

Now the ground is shifting. AI assistants and large language models (LLMs) are pushing us into the Intention Economy. The metric here is not how long a user scrolls, but what they are trying to decide, do, or become.

This shift presents a structural threat. If we allow opaque AI intermediaries to sit between citizens and reality, they will shape human intentions quietly and at scale. Journalism must either step up as essential civic infrastructure in this new economy, or it will be downgraded to raw material for systems that do not share its public mission.

The Pivot: From Attention to Decision

The economic shift is moving from attention to intention. Previously, value came from buying visibility and treating time-on-screen as a proxy for impact. Now, value concentrates in whoever can best predict and shape the user’s eventual decision. AI agents, plugged into everyday queries and habits, turn that predictability into a product.

While AI agents excel at low-stakes intentions—planning a trip or organizing a calendar—they are structurally unsuited for high-stakes uncertainty. This is where journalism’s value proposition lies.

When a user asks, “Who should I vote for?” or “Which career path protects me from automation?”, they are not looking for a generated aggregate of the internet. They are looking for verification. Previously, journalism held the role of mediating reality and scrutiny. Today, that mediator is increasingly a black-box model, trained on everything but accountable to no one.

The Threat: The Unregulated Editor

Answer engines are not merely “better search”; they are unregulated editors of reality.

  • Confirmation, Not Challenge: LLMs are designed to be agreeable. They mirror and extend the user’s assumptions. If a user expresses deep partisan bias, the model is likely to amplify that bias in a plausible, conversational tone—the antithesis of the journalistic duty to challenge perceptions.
  • Structural Hallucination: LLMs predict probable word sequences, not facts. We have already seen systems invent legal precedents or fabricate quotes in political controversies. Unlike journalism, which has explicit and public correction mechanisms, generic AI systems offer almost no transparent way to correct or contest their mistakes. They deliver plausible falsehoods precisely in the contexts where accuracy matters most.
  • The Invisible Editor-in-Chief: When an AI assistant answers a civic question, the user never sees what it omitted or which sources it favored. The AI acts as an invisible editor-in-chief for millions, operating without public scrutiny or civic obligations.

The Response: Decision-Grade Intelligence

If intention is the new battleground, journalism must leave behind engagement-driven filler and commit to Decision-Grade Civic Intelligence.

  • High-Utility, Non-Commoditized Insight: Newsrooms must produce information that AI cannot safely improvise. This includes field reporting on corruption, verified voting records, and specific details on who wins and loses in new policy. This is the kind of insight that retains real economic and civic value.
  • Direct, Accountable Relationships: In an AI-mediated world, maintaining direct, branded channels—newsletters, apps, memberships—is already a political act. First-party data shouldn’t just optimize marketing; it should help newsrooms understand the civic questions their community is struggling with. The goal is that when a crisis hits, citizens turn to a trusted source for verification, rather than settling for a synthetic summary that flattens the truth.
  • Collective Negotiation: We must be realistic about the power dynamics. This is an asymmetrical fight; tech giants control the browsers and operating systems. Individual publishers have zero leverage. The industry must negotiate collectively to establish red lines: mandatory attribution, fair licensing, and a ban on using journalistic work to train synthetic content that competes with the original.

The Choice: Civic Infrastructure or Background Noise

The transition will be brutal. Many publishers will drift into a muddy middle ground—trying to appease algorithms while claiming independence—but that path leads to irrelevance. The only sustainable strategy is to treat journalism explicitly as civic infrastructure.

This requires a dual evolution in distribution and content.

First, we must build a distinguished distribution channel. Journalism cannot survive as generic web feed or sit inside the same undifferentiated stream as influencers and synthetic “slop”. Publishers need to move toward trusted environments defined by clear standards: verified apps and protocols where identity, source provenance, and editorial responsibility are visible by design. The user experience must be fast and mobile-native, but with one non-negotiable: it must be obvious who is speaking, and under which rules. This means choosing ethical tech partners, aligning on common signals of credibility, and refusing integrations that erase brand, byline, or context.

Second, the content inside that channel must adapt. We must abandon the default logic of the feed. The current industry obsession with “streaming news”—endless tickers, alerts, and aggressive push notifications—often feels less like a service and more like harassment. It shreds reality into fragments, offering scenes without stories and emotional spikes without a thread. Newsrooms should restrict real-time coverage to the rare, high-stakes events where users explicitly demand it. For everything else, the primary product must shift from ephemeral updates to stable context: living explainers, timelines, and dossiers that pull the fragments back into a coherent picture.

Ultimately, we must remember why the news exists. The human need for journalism is not about fighting boredom or scrolling through anxiety. It is the instinct to look over the horizon. People need to see what is coming, to understand the forces shifting the ground beneath them, and to prepare for the future before it falls upon them. If journalism settles for being a source of distraction, AI will replace it. But if it acts as the radar that allows citizens to anticipate and navigate the shocks of the future, it will remain the one thing an algorithm can never replace: civic infrastructure.