We tend to treat misinformation as a content problem. Something false enters circulation, spreads, and ideally gets corrected. The solution, in that framing, is better detection, faster fact-checking, and more media literacy. That framing is now outdated. What’s emerging isn’t a flood of false content inside an otherwise stable information environment. It’s the destabilisation of the information environment itself.
The Scale Has Changed the Category
The volume shift is the first thing to understand. According to a European Union report cited in Science magazine, 27% of attempts by foreign powers to manipulate information globally involved AI in 2025 — nearly three times the figure from 2024. Researchers at the ACM Web Conference 2026 describe the result as “industrialised deception”: the automated production of misleading content at a scale no human operation could sustain.
This isn’t just more of the same. When fabrication scales to industrial output, the relationship between misinformation and the information environment changes structurally. Individual pieces of false content can be debunked. An information ecosystem in which fabricated content circulates continuously, at volume, across multiple platforms, in multiple languages, cannot be fact-checked into stability. The UN’s April 2026 brief on information integrity makes the same point: the first UN Global Risk Report ranked misinformation and disinformation as a top global vulnerability — not a communication challenge, but a systemic risk the international community remains insufficiently prepared for.
Three Layers, Not One
The conventional picture of misinformation shows a false claim entering a distribution channel and reaching an audience. What actually exists now operates across three layers simultaneously.
At the content layer, AI generates convincing text, audio, images, and video faster than any detection system can match. The Russian disinformation campaign Matryoshka launched at least 135 pieces of content in the first three months of 2025. One fake video claiming US foreign aid paid actors to visit Ukraine went viral after amplification by high-follower accounts. The other 134 mostly failed. The strategy doesn’t require most of it to work — it requires enough volume that something always does.
At the distribution layer, the architecture of social media platforms does the work that editors and broadcasters once did. Algorithms reward speed, emotion, and shareability. Accuracy is not a ranking signal. The Full Fact 2026 report documents how platform business models actively monetise the spread of misleading content — not through deliberate complicity, but through design choices that generate engagement from outrage and novelty, regardless of whether the underlying content is true.
At the interpretation layer, the cumulative effect of years of this is a generalised distrust that now affects credible sources as much as fabricated ones. When everything feels potentially false, the cognitive cost of assessing any individual claim rises. Many people stop trying. The Social Market Foundation’s work on information access identifies the emergence of local trust networks as a response: people increasingly rely on friends, family, and community groups rather than media institutions, because the latter feel less reliable. This is both understandable and corrosive — localised trust is easy to exploit, and harder to correct from outside.
What AI Actually Changed
Researchers at Indiana University studying AI’s role in misinformation describe the core shift clearly: AI hasn’t just made fabrication cheaper. It has made “plausible truth” cheap. A synthetic video doesn’t need to be indistinguishable from real footage to do damage. It needs to be plausible enough that viewers doubt themselves rather than the content.
This is what’s different from the pre-AI era. The cost of creating a convincing fabrication used to act as a natural filter. Producing a fake video of a politician required production capacity, distribution infrastructure, and some plausible cover story. Now it requires a laptop and a few minutes. The ACM 2026 research frames this directly: LLMs have eroded trust not only in specific content but in the information infrastructure itself.
That phrase — information infrastructure — is worth dwelling on. Infrastructure is what you depend on without thinking about it. Water, electricity, roads: you assume they function until they don’t. The information infrastructure that democratic societies depend on — shared reference points, reliable sources, correction mechanisms when false claims circulate — is now under sustained structural pressure. As the Full Fact report puts it, public interest information is “essential for stabilising the information environment,” but that infrastructure itself faces sustained pressure.
The Correction Problem
One of the most consistent findings in misinformation research concerns the asymmetry of impression and correction. A false claim generates an impression. The correction almost never fully undoes it. In some cases — the backfire effect, though its magnitude is debated — correction entrenches belief rather than reversing it.
This means that even a perfectly functioning fact-checking infrastructure would not solve the problem. The solution to a content problem is better content. The solution to an infrastructure problem requires working on the architecture, the incentives, and the institutional structures that govern how information flows — not just on the claims circulating within it.
This connects to the pattern explored in When Systems Decide for Us, What Happens to Human Judgment?: as AI mediates more of what people see and believe, the capacity to exercise independent judgment about the reliability of what you’re receiving becomes both more important and harder to develop. The same dynamic operating in health tracking and financial advice operates in the information environment — systems that optimise for something other than truth, at scale, reshape what people believe without anyone making a single decision to deceive.
Infrastructure Requires Infrastructure-Level Responses
The UN brief is clear on what this demands. Advertising standards that prevent platforms from monetising misinformation. Platform design accountability. Investment in public interest information as a public good. Meaningful AI literacy as an educational priority, not a supplementary skill.
None of these are content interventions. They’re structural ones. The shift from treating misinformation as a content problem to treating it as an infrastructure problem is the shift that needs to happen in policy, in platform design, and in how institutions invest in the information environment they depend on.
We are no longer dealing with false information circulating inside a stable system. We are dealing with a system that is itself becoming unreliable — and that distinction makes almost everything that followed from the old framing insufficient.
Key Sources
- UN Information Integrity Issue Brief – Strengthening Information Integrity, April 2026
- Full Fact – Full Fact Report 2026
- Science / AAAS – To Misinformation Researchers, AI Is a Scourge — and a Powerful New Tool
- ACM Web Conference 2026 – Industrialised Deception: The Collateral Effects of LLM-Generated Misinformation on Digital Ecosystems
- Social Market Foundation – Information Access and Local Trust Networks
- Frontiers in AI – AI-Driven Disinformation: Policy Recommendations for Democratic Resilience
Subscribe to EuroLuminant for independent European journalism.
