Information manipulation and online harm across the EU
How Europe's information space is targeted in 2026, by whom, on what infrastructure, and what platforms and governments did about it.
November 2026 · Drawing on Alliance4Europe’s and members of the Counter Disinformation Network’s research published January to November 2026.
Europe's information space is attacked on an ongoing basis by known influence operations, on some of the same infrastructure that was exposed back in 2024 and 2025. What has changed is the speed, the use of frontier AI models to produce content at scale, and how little it now costs to keep going. One network has moved across four elections; Russia's Matryoshka and Doppelganger operations keep running despite sanctions; and platforms keep paying the operators of political slop. This is a security problem, not only a media one: the same operations target elections, relations between allies, trust in crisis institutions and support for defending Ukraine and Europe. They are expected to accelerate around Europe's key elections in 2027. What’s needed is investment in resilience.
Detection is not the problem. Prevention, resilience, response, and enforcement are.
Key findings
Each finding ends with tags. [Data] marks what was measured; [Analysis] marks our assessment of it. Numbers in brackets point to the Notes.
- One network, four elections. An X network of more than 550 fake accounts exposed in the Dutch election (October 2025) switched to Hungary's April 2026 election and then to Sweden, where we identified 8,505 accounts in August. Two accounts are already following Serbian accounts ahead of Serbia's early presidential election, due by 27 December. As far as we can assess, X has removed fewer than 1% of the accounts logged earlier. [Data: §1, Figure 2] [Analysis: §7] [6, 7]
- Russian operations run on, despite sanctions and exposure. On 18 August, less than a month before Sweden's vote, Matryoshka published at least six fake news videos branded as six European outlets. X declined our notice and removed the content six days after publication; Bluesky disrupted the same operation quickly. Doppelganger returned on 7–9 October with clones held in reserve for 17 months, targeting the US midterms, Israel's election and France's 2027 race at once. [Data: §2, Figure 3] [2, 8]
- Domestic anger, foreign amplification. France's October–November protests were home-grown, with no foreign fingerprint in hosting, domain or tracking records. Foreign actors worked the same grievances: 639 Doppelganger messages before the protests, 4,299 Pravda network articles across 55 sites in 17 languages, and a quarter to a third of Iranian state media's French-language output. Abroad, 13.8% of English-language posts blamed migrants or Islam, against 2.9% of French posts. [Data: §3] [3]
- AI is now the target, the tool and the referee. 29 YouTube channels mass-produced more than 7,300 AI-narrated geopolitical videos; one Hungarian Facebook page kept five AI political videos running for 50 days through 121 ads; 46 of 100 chatbot answers in a ten-chatbot audit went along with disinformation narratives about Lithuania; 461 posts asked X's Grok to rule on what was true during the French unrest. [Data: §4] [3, 9, 10, 11]
- Domestic actors are borrowing the playbook. The day before Sweden's general election, the Sweden Democrats sent text messages that, in practice, targeted voters by ethnic or religious background. Data-protection experts assess this as a likely GDPR breach; we classify it as a domestic influence operation. [Data: §5] [Analysis: §5] [12]
- Disinformation is also a business. Reset Tech tracked 6,601 accounts posting 962,000 AI-generated political clickbait items to more than 4 billion views. 84% of Facebook pages demonetised for repeated disinformation regained monetisation, and 100 hate-speech channels are on track to earn about $13.3 million on YouTube this year, with YouTube keeping about $6 million. [Data: §6] [14, 16, 17]
- The enforcement gap is the problem. Google removed none of 15 Hungarian political ads identical to ones Meta had removed; TikTok replied "No violations" to a 105-account coordinated network; Meta removed 73% of 1,739 reported items, but ad-by-ad removal let one campaign run for 50 days. [Data: §7] [Analysis: §7] [2, 7, 10]


