Authors: David Hobson, Jean-Romain Roy, Ben Steel, and Aengus Bridgman

Organization: Network Dynamics Lab and Media Ecosystem Observatory

Key takeaways:

There has been little direct public digital interaction between Canadian politicians of any political party and the Tenet Media personalities over the past two years and no Canadian politicians are consistently producing content similar to that of the Tenet Media personalities.
However, there is a set of Canadian influencers who do directly interact with the Tenet Media personalities and who produce content that aligns in substance and timing. These influencers serve as a bridge between the Tenet Media personalities and the larger Canadian information ecosystem.
The content of Tenet Media personalities tends to align to the topic areas of interest to the Russian funders and also outperforms posts about other subjects. Canadian influencers also post content that aligns with Russian objectives but these posts are less frequent and tend to only somewhat outperform other topic areas.

As part of our ongoing investigation into the funding of Tenet Media by Russian intermediaries promoting specific topics and narratives, we assess the relevance of influencers hired by Tenet Media to the Canadian information ecosystem. Here we focus on all those hired by Tenet Media — five American political commentators Benny Johnson, Dave Rubin, Tim Pool, Matt Christiansen, and Tayler Hansen, and one Canadian political commentator Lauren Southern (hereafter collectively referred to as Tenet Media personalities).

All six influencers have denied being aware that any of the money came from Russian sources.(1) In our investigation we also observed that Tenet Media personalities post only infrequently about the Russia-Ukraine conflict. We make no claims about awareness or complicity, and instead focus on the key fact that the Russian funders are targeting these individuals, their audience, and their broader online community or neighbourhood.

The Russian interest in this community is material and an examination of the Canadian implications is key to understanding who Russia is ultimately interested in encouraging and influencing. Throughout this incident update, we have chosen to not include the names of the Canadian influencers and politicians in the network plots to not polarize and distract from the central findings of the update.

Do Canadians interact with Tenet Media personalities?

We begin the investigation by looking at which Canadian politicians and influencers (hereafter Canadian entities) have direct interactions with the Tenet Media personalities. Specifically we look at which Canadian politicians and influencers have been publicly engaging with their content, and which Canadian politicians and influencers are the Tenet Media influencers publicly engaging with. For this investigation, we focus on initiated interactions (i.e. any mention of, report, reply, or quote on 𝕏 , or an Instagram co-authorship or usertag) between Tenet Media personalities and the Canadian Digital Media Research Network’s list of Canadian entities. Over the period examined, Tenet Media personalities initiated an interaction with 55 distinct Canadian entities a total of 715 times, and 161 distinct Canadian entities initiated an interaction with the Tenet Media personalities a total of 1993 times. We visualize these direct interactions as networks in Figures 1 and 2.

Figure 1 shows interactions initiated from Tenet Media personalities (large yellow circles) to Canadian entities (shown as medium circles and referred to as those with one degree of separation) and then initiated interactions from those to other Canadian entities (shown as small circles and referred to as those with two degrees of separation). Example 1 shows an initiated interaction from July 2024 where Matt Christiansen mentions @BillboardChris - a BC-based political activist with a large audience. Example 2 shows an initiated interaction from June 2024 where Dave Rubin co-publishes a conversation with Gad Saad, a Professor from Concordia University in Montreal who has a very large online following, on Instagram.

Example 1: A Tenet Media personality mention of an influential Canadian entity on

Example 2: A Tenet Media personality co-posting with an influential Canadian entity on an Instagram post

In Figure 1, Canadian influencers are highlighted in grey with Canadian federal and provincial politicians in Liberal red, Conservative blue or NDP orange. The relative placement of each entity is determined by frequency of interactions, with a force-directed algorithm positioning entities closer together when they share many common interactions. A one degree of separation means at least one interaction initiated by a Tenet Media personality to a Canadian entity. A second degree of separation means a Tenet Media personality initiated an interaction with a Canadian entity (the one degree) who in turn initiated an interaction with another Canadian entity (the second degree).

Figure 1: Network visualization of directed interactions from Tenet Media personalities to Canadian entities within 2 degrees of separation.

Critically, Tenet Media personalities have initiated interactions with only a small number of politicians in Canada (Justin Trudeau, Pierre Poilievre, Maxime Bernier, Melissa Lantsman, Mark Gerretsen, Francois-Philippe Champagne, and Sean Fraser, shown with examples of interactions). Instead, the bulk of their interactions are with Canadian influencers who in turn interact with other Canadian entities. In this way, Canadian influencers provide a bridge between Tenet Media personalities and the broader Canadian information ecosystem.

Next, in Figure 2, we look at interactions initiated to Tenet Media personalities (large yellow circles) from Canadian entities (shown as medium circles and referred to as those with one degree of separation) and then from other Canadian entities to those one degree of separation entities (shown as small circles and referred to as those with two degrees of separation). Example 3 shows a one degree of separation interaction from Jordan Peterson, a former professor from University of Toronto with an enormous online following to Tenet Media personality Lauren Southern on 𝕏.

Example 3: A Canadian account on 𝕏 quote tweeting a Tenet Media personality

Figure 2: Network visualization of directed interactions from Canadian entities to Tenet Media personalities within 2 degrees of separation

Like Figure 1, Canadian influencers are highlighted in grey with popular online profiles of Canadian politicians in Liberal red, Conservative blue or NDP orange. Again, positionality in the network is determined using a force-directed algorithm, with entities positioned close together being those who share common interactions. There are only five instances of a politician directly interacting with a Tenet Media personality - all relatively minor. (2) The bulk of interactions are with Canadian influencers who in turn are interacted with by other Canadian entities. Again, it is a set of Canadian influencers that provide a bridge between Tenet Media personalities and the broader Canadian information ecosystem.

What influence do Tenet Media personalities have in Canada?

This influencer bridge appears to be a key conduit by which the messaging and content of interest to Russian operatives enters the Canadian information ecosystem. We next map the influence Tenet Media personalities have in Canada by evaluating the similarity between what Tenet Media personalities and Canadian entities have posted over the past two years. For every possible pair of entities (e.g. Benny Johnson to Justin Trudeau, Justin Trudeau to Pierre Poilievre), we assess the directional influence of one entity to another through behaviours. Specifically, we look at how often one entity shared the same web links, used the same hashtags, mentioned the same other users, or posted semantically similar content, to another. We aggregate each of these behaviours and combine them into a single directed measure that captures the strength of the influence from one entity to another. We then take all those connections and project them into a network visualization using a force-directed algorithm: entities are pulled closer together when they influence one another (share a lot of similar content and behaviors) and pushed further apart where they do not.

Figure 3 shows a projection of these connections, with each circle representing the social media accounts of a Canadian entity, each size of circle representing how much engagement they receive online, the color of the circle representing the type of entity (i.e. Conservative politician, Liberal politician, Canadian influencer, etc.), and the distance and clustering between the circles indicating the similarity of their content in comparison to other personalities and entities. See the methodology for additional details. The Figure also shows the neighbourhood (an online community that posts and behaves similarly) of the Tenet Media personalities in yellow. The Figure visually shows circles of Canadian influencers that reside in the Tenet Media neighbourhood as well as Canadian politicians (not in the neighbourhood) as fully opaque, while other Canadian entities are faded out.

Figure 3: A directed network visualization of Tenet Media personality influence and proximity to Canadian entities

Again, we do not see a single Canadian politician at either the federal or provincial levels in the neighbourhood of the Tenet Media personalities. This finding is critical - no Canadian politicians are consistently producing content similar to that of the Tenet Media personalities. Instead, there are approximately 20 Canadian influencers who are within their neighbourhood and produce similar content. Some of these influencers are also those who mention and are mentioned by them. The impact that Tenet Media personalities have in the Canadian information ecosystem is partially through these Canadian influencers. In effect, the yellow zone highlighted in Figure 3 indicates the general community target of the Russian operatives, with the rest of the information ecosystem subject to the influence of those contained within.

What do these bridge influencers focus on?

What are these influencers discussing? In IU4 which focused on the podcasts of the Tenet Media personalities, we evaluated the topics that Tenet Media personalities focused on in relation to Canada and the stated topics/narratives that the US-indictment flags as those Russian operatives want to “exploit in the course of an information campaign in/for the United States.” Here, exploit refers to the objective of exacerbating the already-high levels of political polarization in the United States, which is also on the rise in Canada. They are as follows (with quotes from the indictment):

The economy and inflation (“Encroaching universal poverty. Record inflation. Halting of economic growth. Unaffordable prices for food and essential good.”)

Job loss (“Risk of job loss for white Americans.”)
Issues of social justice focused on race, gender identity, sexual orientation, and disability rights (“Privileges for people of color, perverts, and disabled”)
Controversies and/or misinformation (“Constant lies of the [Biden administration] about the real situation in the US”)
Migrants/Immigration (“Threat of crime coming from people of color and immigrants (including new immigrants from Ukraine)”)
Foreign spending (“Overspending on foreign policy and at the expense of interests of white US citizens”)
Amplifying false claims (“Constant lies to the voters by the [Democratic Party] in power.”)
Escalating war (“America is suffering a defeat despite [Biden’s] efforts. We are being drawn into the war. Our guys will die in Ukraine.”)

We use a similar approach to IU4 to use large language models to classify all the posts on 𝕏 of: 1) all the Canadian bridge influencers (those in the shaded yellow neighbourhood in Figure 3); 2) a random sample of high-engagement posts across all other influential Canadian entities; and 3) all posts from Tenet Media personalities. All data is from January 1, 2023 to September 19, 2024, as given previously. Given the low percentages in some of the categories, we grouped several of the topics together:

Unjust war/overspending on foreign interests (Topics 6 + 8)
Social justice (Topic 3)
Inflation, cost of living and unemployment (Topics 1 + 2)
Crime and immigration (Topic 5)
The Biden administration (Topics 4 + 7)

Figure 4 is a matrix plot showing the frequency these topics are discussed and engaged with for each Tenet Media personality - topics are plotted along the y-axis and personalities along the x-axis. Two Canadian groups of accounts (bridge influencers and all other influential Canadian entities) are also included on the right. The upper triangle for each entity-topic pair shows the percentage of all content shared that is about that topic and the bottom triangle shows the percentage of overall engagement with all content for the topic-entity combination. For example, the top left box shows that 40% of Benny Johnson posts are about the Biden administration and those posts receive 40% of his overall engagement. Triangles with higher percentages are darkened to show importance. For example, 11% of Lauren Southern’s posts are on issues of social justice but 21% of her engagement comes from those posts (i.e. they are comparatively much more popular). Note that all percentages are estimates due to model error (the classification technique being necessarily imprecise) and results should be interpreted as directional. Percentages do not add up to 100 because not all posts align with one of the narratives.

Figure 4: Topic frequency and proportion engagement aligned with stated Russian objectives for Tenet Media personalities.

A significant proportion of the content generated by The Tenet Media personalities aligns with the five topics, ranging from around 25% (Tayler Hansen) to over 50% (Benny Johnson). In all cases, this content overperforms in engagement: their audiences are much more likely to engage with content related to these five topics than other topics. All the American Tenet Media personalities frequently post critiques of the Biden administration, often in a controversial way (e.g. from Benny Johnson: “WATERGATE 2.0! Biden ORDERED FBI Raid on Trump as STUNNED White House Lawyers Declare Raid ILLEGAL”), and also posts frequently about crime/immigration and social justice. The one Canadian personality is much less focused on American political scandal and instead posts heavily about social justice issues. She also mentions inflation and the economy at a higher rate than the Americans - this is primarily in the context of the housing crisis in Canada.

Example 4: A post from Tenet Media personality Lauren Southern which has been classified as about Crime and immigration

The Canadian bridge influencers do post more about these topics, overall, than other Canadians. This is driven chiefly by a much higher interest in American politics than the average Canadian entity account. However, there is no substantial difference in engagement on other topics between the bridge influencers and the other Canadian accounts. This suggests that the specific topics of interest to RT operatives are of lesser importance to the Canadian conversation. This is in stark contrast to the Tenet Media personalities who both post more about the topics, but also see much larger engagement from them. Overall, these bridge influencers are more tuned in to American politics and pay closer attention to those issues but are otherwise not producing content aligned with the stated Russian operatives objectives.

Methodology

Direct interactions

Data for the direct interactions as well as the similarity network is from the Canadian Digital Media Research Network dataset of social media activity of Canadian politically influential entities, which includes relevant metadata (e.g. likes, shares, etc.). We counted a direct interaction as any of: a mention of another user on any platform, an 𝕏 retweet, reply or quote tweet, or an Instagram co-author and usertag. Looking at the range of January 2023 to September 2024, we summed all these interaction counts between users with equal weighting for each type. Due to high interaction counts between some users, we applied the log function to all summed interaction counts, with the intuition that there is a diminishing importance as interaction numbers increase.

We use these logged interaction counts to construct a network. Using Djikstra’s algorithm to find shortest path lengths, we find the set of nodes that are at most two degrees of separation from the Tenet Media personalities, by interactions targeted from the Tenet Media personalities to Canadian public figures for Figure 1, and by interactions targeted towards the Tenet Media personalities for Figure 2. Filtering the network to only these close public figures, we then combine directed edges into a single undirected edge between public figures by the harmonic mean, with the intuition that interactions between two figures are more heavily weighted downwards by the weaker relationship. Put another way, if person A frequently interacts with person B, but person B only occasionally interacts with person A, the mutual relationship weight is only slightly higher than the lowest weight. We then apply the Force Atlas 2 network layout algorithm to the network, from the cugraph Python library, to obtain two dimensional node positions. We group the nodes into 6 groups: Liberal politicians, Conservative politicians, NDP politicians, politicians of any other party, Canadian influencers, and Tenet Media personalities. We size the node based on its degree of separation from the Tenet Media personalities.

Similarity network

To analyze the influence of Tenet Media personalities on Canadian political discourse, we assessed the connections and content similarities between Tenet Media personalities and Canadian entities. Using a list of 2,254 Canadian politicians and prominent Canadian entities, we measured the level of alignment between Tenet Media personalities and Canadian entities based on shared web links, hashtags, mentions, and simultaneous content themes.
Entities with one or more social media accounts served as nodes in the network (individual entities represented as circles), and similarity in content (links, hashtags, mentions, and textual similarity within a 24-hour period for post pairs, shown as lines in the Figure) between accounts formed the edges. We calculated an edge weight based on the frequency and relevance of shared content, filtered by an edge threshold to focus on the strongest connections. Node size was scaled by the log of total engagement, and nodes were categorized by political affiliation (e.g., Liberal, Conservative) or influencer type, with Tenet Media personalities marked as a distinct class.
We constructed an undirected network in R using the igraph package, where each node represents an influencer or politician, and edge weights represent content similarity. The Fruchterman-Reingold force-directed algorithm arranged the network layout, with nodes pulled closer for high similarity and pushed apart for low similarity, enabling clusters of related entities to form naturally.
The network projection displays Canadian entities, with each node’s size reflecting the log of total engagement and color indicating type (e.g., Conservative politician, Liberal politician, Canadian influencer). Nodes within the Tenet Media sphere of influence and important Canadian politicians appear fully opaque, while other entities are faded. The distance between nodes in the final network plot represents the level of content alignment. Closer nodes signify greater content similarity while farther nodes indicate dissimilarity.
The density region surrounding Tenet Media accounts is shaded. To visually highlight the neighbourhood of Tenet Media, we conducted a 2D density estimation (kde2d) within the network to identify high-density clusters of accounts that shared similar content with Tenet Media personalities. We set a cutoff threshold to emphasize regions with strong similarities in content.

Topic classification

We first collected the posts from all the Tenet Media personalities: @Lauren_Southern, @bennyjohnson, @RubinReport, @TaylerUSA, @Timcast, @MLChristiansen, for a total of 33,682 posts from January 1, 2023 to September 19, 2024. We then used GPT-4o to classify the topics of the tweets using the prompt below. Topics 6 and 8 were combined (into Unjust war/overspending on foreign interests), topics 1 and 2 were combined (into Inflation, cost of living and unemployment), and topics 4 and 7 were also combined (into the Biden administration). 23, 458 posts from bridge influencers identified in the network analysis were also classified using the same method for the same time period. Finally, 21,075 posts from Canadian entities were also classified using the same method.

***

For the following text passage, indicate which of the following categories best applies to the text. Indicate your answer as a single number with no explanation. If more than one category applies, choose the most applicable one, and if two apply equally, select either number. If none of the categories applies, reply with "0".

Categories:

Inflation, halting of economic growth, unaffordable prices for food and essential goods;
Job losses (or risk of job losses) for white Americans;
Privileges or general discussion about people of color, the disabled, or the LGBT community;
Threat of crime coming immigrants or from people of color;
Overspending on foreign policy, especially at the expense of white Americans;
That the US is being drawn into the war in Ukraine.
Lies about the Biden administration;

Text:"""

{text}

"""

Footnotes

(1) Dave Rubin, Tim Pool, Lauren Southern, Benny Johnson, Tayler Hansen, and Matt Christiansen.

(2)https://x.com/johnkoury/status/1749671794657968162, https://x.com/bobbaileypc/status/1813732416412086654, https://x.com/bobbaileypc/status/1827699607842755058, https://x.com/fsorbara/status/1730788967409238412, https://x.com/CHPCanadaLeader/status/1753488251980243216, https://x.com/Prominent_Bryan/status/1794924563224236104