When Machines Read The news

We often discuss disinformation as if it were merely a content problem.

We look at it as a plague of state actors, fake stories spread by hostile powers, deepfakes produced by criminals or greedy influencers gone rogue. That framing is no longer sufficient. Disinformation is no longer primarily about what is being said. In 2026, it is about how the information becomes visible, how it is being processed, ranked and reused. In other words: it's a systemic problem.

For most of modern history, news was written by people, for people. Context was interpreted by humans and trust was relational: between journalists, institutions and their audiences.That world is rapidly disappearing. Today, the news is increasingly processed machine-first: always ranked by algorithms, often summarised by AI systems and most of all redistributed by platforms optimised for engagement. Currently, there are likely more automated clicks on news sites that human ones.

As a result, the power to influence public opinion has shifted: from persuasion to machine ingestion, from editorial judgement to data structure, and from authority to algorithmic visibility. What circulates is no longer what the public consciously wants to read. It is what systems select to satisfy their business logic.

This creates an uncomfortable tension: technology platforms effectively shape what is visible, viral, and credible yet they cannot, and should not act as arbiters of truth. Their systems are optimised for engagement, retention and, advertising revenue. Extreme, emotionally charged content performs better than nuance or context, and AI tools dramatically lower the cost of producing such content at scale.

The problem is not just “echo chambers”. It is something more dangerous: self-reinforcing information systems that amplify extremes, generate synthetic content, and then re-consume that same content as training data. Effectively, these systems are slowly poisoning themselves.

At the same time, it would be a mistake to assume that journalism and traditional newsrooms, or social media influencers can solve this alone. Journalists no longer control distribution. Editors no longer set the flow of information. Influencers only exist by the grace of the algorithm. Whether you are a journalist, an influencer or an engaged citizen, you are now a micro-distributor inside a macro-system you do not control. The classical role of journalism as the “fourth estate” is not failing because of quality or intent. It is being structurally undermined by algorithmic systems that operate at a different scale.

Private initiatives, however well-intentioned, will fail if left to compete in isolation against platform-scale dynamics. But the alternative of state control over information is equally unacceptable. We risk getting caught between two fake solutions: private optimisation by tech giants without public responsibility, or public control with unacceptable risks to freedom and pluralism.

What we are missing is something else. Healthy systems usualy rely on reference layers: financial markets rely on reference data, aviation on air-traffic control or science on peer review. Our information ecosystem needs a comparable, neutral reference layer. This is not about opinion, framing or narrative control. Instead, it requires verifiable, attributable, and machine-readable facts that serve as stable inputs for both humans and systems.

Without that, trust erodes. Not because truth disappears,but because it becomes structurally unfindable. Once information is consumed, ranked and reused by machines, integrity becomes a matter of public resilience. It impacts democratic legitimacy, social cohesion, public safety and economic stability.

That is why the solution cannot be reduced to media literacy, platform moderation, AI governance or cybersecurity alone. Those are necessary, but insufficient. If disinformation is a system failure, then the response must be systemic. We need to restore a neutral, professional reference layer in our information systems. This layer must be independent, data-driven, politically neutral and designed for both human and machine use. It exists not to dictate narratives or control opinions, but to anchor facts, attribution and corrections in a way that systems can reliably ingest and reuse.

In a world where machines increasingly decide what we see, whoever stabilises machine-readable truth stabilises society itself. The real question is whether the humans responsible for those systems can trust the sources they feed into the machines. That trust cannot be based on algorithmic probabililty. It must be grounded in how information is produced, verified, attributed and corrected.

This is where "human in, human out" becomes more than an ethical slogan. It becomes an operational guarantee. When information is produced with rigour, machine ingestion becomes a safe operational choice. AI systems can then use these structured, verifiable source to cross-check claims in real time and resolve contradictions instead of amplifying them. This allows them to clean their own training pipelines, reducing the recursive poisoning that currently undermines their reliability. In this model, machines do not replace human judgment. They scale it, without eroding trust.

Let's conclude with an uncomfortable question.

If your strategy relies on mastering communication tactics, optimising for algorithms, courting influencers, gaming visibility and chasing AEO, ask yourself this: where will you anchor the underlying facts?  And what happens when the system turns against you?

Systems optimised for engagement do not distinguish between ally and target. They amplify whatever performs.When misinformation hits you, when your reputation is distorted by synthetic narratives, selective disclosures or algorithmic noise, where will you turn to recover the truth? Who will you trust to re-establish reality, if everyone is spinning inside the same machine-driven loops?

My advise? Follow the example of Odysseus. Listen to the Sirens if you must. Their song is irresistibly optimised; you cannot silence them, and you cannot out-sing them. But don't forget to tie yourself to the mast.

© Belgaimage

This blog is written by Patrick Lacroix in a personal capacity. AI tools are used for research, structuring, drafting and language support. All content is selected, verified, and edited by the author, who retains full editorial responsibility.