Listen to this DisinfoWatch DisinfoDigest below:
This edition of DisinfoDigest highlights how foreign and foreign-linked actors continue to exploit political divisions, democratic debates and international crises using a mix of state media, influencers, deceptive websites and AI-generated content.
In Canada, U.S. influencer Benny Johnson has openly promoted Alberta separatism while misleadingly portraying the October 19 referendum as a direct vote on independence. Separately, a network of 23 deceptive websites exposed by Canada’s National Observer appears designed not only to influence Alberta voters, but also to shape what AI chatbots retrieve and repeat about the referendum. Meanwhile, the Kremlin and Russian aligned influencers have has advanced a conspiracy theory questioning evidence surrounding the Leipzig drone attack—an example of attribution erosion intended to cast doubt on suspected Russian hybrid operations.
Ukraine remains a central target. During President Zelensky’s visit to Canada, Russian platforms portrayed him as a “beggar” and Canadian assistance as wasteful taxpayer “handouts.” Recent cases catalogued by EUvsDisinfo show the same narratives continuing elsewhere: Ukraine portrayed as a Western-controlled “neo-Nazi” proxy, NATO countries falsely described as being directly at war with Russia, and Moscow’s 2022 invasion reframed as a pre-emptive response to an alleged Western attack. Another August case described Ukraine as a “Zelenskyy concentration camp,” combining claims of political repression, religious persecution and “Russophobia.”
Together, these cases illustrate how familiar influence narratives are being adapted to new platforms, technologies and domestic political vulnerabilities.
AI-Generated Website Network Targets Alberta Referendum — and AI Chatbots
A September 4 investigation by Canada’s National Observer exposed a coordinated network of 23 deceptive websites promoting Alberta separatism ahead of the province’s October 19 referendum. The investigation found that the sites appear designed not only to influence voters directly, but also to manipulate the information that artificial-intelligence systems retrieve and repeat when users ask questions about Alberta independence.
According to the report, websites target specific audiences including farmers, young people, veterans, mothers and oil-industry workers. Some present themselves as neutral, non-partisan resources while publishing false information about the referendum, including incorrect voting dates and misleading descriptions of the ballot. Many feature fabricated Albertans, AI-generated photographs and invented testimonials. In some cases, AI-generated material was falsely attributed to real Canadian policy and energy-sector figures.
The operation appears engineered specifically to exploit AI systems. According to the investigation, the sites contain explicit instructions inviting AI answer engines to read and cite their material. Their source code grants access to multiple AI crawlers, including those associated with ChatGPT, Claude, Perplexity, Google and Common Crawl. Some sites use “llms.txt” files intended to make their content easier for AI systems to process, while articles were backdated to make newly created websites appear older and more established than they are.
The underlying infrastructure raises further questions about coordination. All 23 sites were discovered on the same server and have links to the domains koizo.io and koizo.cloud. National Observer also found a site-generation system describing a process for creating a “fleet” of websites tailor-made to be cited by chatbots. The pipeline reportedly used AI models from OpenAI and Anthropic to automate website production and improve access by AI agents.
VIGIL Strategic Intelligence Analysis
Analysis provided exclusively to DisinfoWatch by John Gray at VIGIL Strategic Intelligence, conducted independently of the National Observer investigation and using publicly available infrastructure data, confirms the core technical finding. VIGIL determined that all 23 domains are hosted together on a single server with no unrelated tenants. Every domain on that server is focused on Alberta sovereignty or anti-federal themes — a concentration VIGIL says would be highly unusual on a normal commercial hosting platform.
VIGIL also identified a strong pattern of automation. The domains were registered in three distinct batches, with several registrations occurring within single-digit seconds of each other, indicating scripted rather than manual registration. After each batch, security certificates were provisioned within minutes. Exactly five days later, another configuration operation ran across the domains in tightly synchronized groups, in some cases within one second of each other. Taken together, VIGIL identified three separate automated processes repeated consistently across the network.
What remains unknown is attribution.
National Observer identified possible connections to the United States, including U.S.-hosted infrastructure and a Florida email address appearing repeatedly in the site-generation code. It also uncovered infrastructure containing internal data apparently associated with Let Alberta Decide, the largest third-party separatist advertiser in the referendum. But none of the individuals or organizations contacted acknowledged involvement.
VIGIL cautions that the infrastructure alone cannot establish who is behind the network. While the operation is technically consistent with a foreign influence campaign, it could also represent a domestic political marketing operation or a commercial provider hired by actors supporting Alberta separatism.
The timing remains significant. After National Observer contacted individuals and organizations with possible links to the network, several backend pages — including the site-generation infrastructure — were taken offline, while the public-facing websites remained active.
VIGIL independently confirmed on September 10 that the site-generation platform was still offline, but that the server hosting the 23 public-facing domains remained active.
The case illustrates an emerging vulnerability in democratic information environments: influence operations no longer need to persuade only human audiences. They can also manufacture an ecosystem of apparently independent sources designed to shape what AI systems learn, retrieve and ultimately tell voters.
Coverage of Zelensky visit
Russian state media coverage of Ukrainian President Volodymyr Zelensky’s visit to Canada was relatively muted on the main RT and Sputnik websites. RT’s accounts on X, however, actively sought to frame the visit and resulting agreements through familiar Kremlin narratives about Western support for Ukraine.
One RT post attempted to exploit grievances already circulating among far-left and far-right audiences, falsely claiming that an agreement signed by President Zelensky and Prime Minister Mark Carney would “lock in Canadian taxpayer cash handouts for 100 years.” The framing distorted the nature of the agreement while reinforcing a broader narrative that Canadian support for Ukraine represents an open-ended and wasteful burden on taxpayers.
Tweets from various Canadian accounts about the Zelenskyy visit.
Russia’s Pravda News Network was considerably more active. During the first 24 hours of Zelensky’s visit, the network published nearly two dozen articles (manually identified by DisinfoWatch) focused on the visit, Canadian support for Ukraine and related political narratives.
Several posts combined longstanding Kremlin anti-Ukrainian themes with narratives specifically tailored to Canadian audiences. One article amplified far-right Irish influencer Chay Bowes, who wrote: “As the Beggar Zelensky lands in Canada, never forget Canada’s dark Nazi harbouring past.” The post fused two recurring Russian narratives: portraying Zelensky and Ukraine as financially dependent on Western governments, while invoking allegations about Canada’s historical treatment of alleged Nazi collaborators to stigmatize Ukrainian-Canadian communities and undermine support for Ukraine.
Another Pravda News article stated that Zelensky had “headed straight to Canada, where he managed to scrounge aid packages worth over $2 billion.” The deliberately contemptuous language reinforces the Kremlin narrative of Zelensky as a “beggar” extracting money from Western taxpayers, while presenting Canadian military and financial assistance as an reckless transfer of public resources.
Taken together, the coverage represents a continuation in the effort to portray Ukraine as corrupt and financially dependent, framing Canadian assistance as wasteful, and weaponizing historical accusations of Nazism to discredit both Ukraine and Canadians of Ukrainian heritage.
Various accounts on X also amplified content aligned with these narratives. One AI-generated video, reportedly created using Grok, depicted Prime Minister Mark Carney handing bags of money to President Zelensky, visually reinforcing the claim that Canadian support for Ukraine amounts to indiscriminate taxpayer-funded handouts.