This report analyses which narratives about Latvia, consistent with the Kremlin’s position, appeared on social media ahead of the elections to the Saeima (Latvia’s parliament) on 3 October 2026, who spread them, and whether inauthentic activity was involved, such as the same text being posted from several accounts. The study is based on monitoring by the Centre for East European Policy Studies (CEEPS) and DisInt.ai, which covered 502 public channels on YouTube, Telegram, TikTok, Facebook and X from 1 July to 4 October 2026.

In the three months before the elections, the volume of these narratives more than doubled, and as election day approached they were increasingly aimed at the elections themselves. Yet most of the narratives were spread not by Kremlin-aligned channels or by accounts showing signs of inauthentic activity, but by ordinary users in comment sections. The data therefore give a fairly clear answer to the question posed in the report’s title: in the 2026 elections, the Kremlin’s voice was quieter than its echo among the domestic audience. That echo is the main challenge in the information space in the pre-election period.

Key findings

Kremlin narratives about Latvia were not new, but they were adapted to the pre-election agenda. They have long rested on a small set of claims – for example, that Latvia cannot exist without Russia, is not a sovereign state, persecutes Russian speakers, provokes war with Russia and honours Nazism. In the pre-election period, these claims were linked to current domestic political events, such as the grain transit tariff, fuel prices and airBaltic’s financial difficulties. The number of mentions rose from an average of 230 a day in July to 584 a day in election week, and roughly half of that increase came in the last two weeks before the elections and on the day after them.

The elections themselves became the main target of the narratives. The fastest-growing claim was that Latvia’s elections are illegitimate. Such a claim undermines trust in the electoral process and may reduce turnout among certain groups of voters. It peaked during the elections: in the first four days of October it appeared 461 times – more than in July and August combined – most often on the day after the vote. Commenters presented the low turnout, the delay in counting the votes and the exit poll results as proof that the outcome had been decided in advance. Calls to restore relations with Russia and Belarus also became more widespread. By contrast, narratives about history and Soviet monuments, which normally make up a large part of the Kremlin’s information flow about Latvia, declined in the pre-election period.

Kremlin narratives about Latvia are spread mainly by the audience, not by content creators. Four out of five recorded instances of a narrative were found in comments, and the largest comment sections are on the accounts of Latvian politicians, parties and media. The two groups of Kremlin-aligned channels together accounted for only 13% of all instances, and their activity remained at a similar level throughout the period. The pre-election rise in narratives therefore came mainly from user comments.

The network of Kremlin-aligned channels is small but well structured, and in the elections it did not line up behind any one list. The core of the network consists of about a dozen Telegram channels that regularly repost each other’s texts. Although Russian state media are restricted in Latvia, the Russian state outlet Sputnik Latvia (Sputnik Латвия) sits at the centre of the network. In the pre-election period, posts by Latvian politicians also spread along these reposting routes, but in election week the most widely reposted material discredited the leader of one pro-Russian list.

Inauthentic activity was limited but targeted. Accounts showing signs of inauthentic activity posted fewer than one in ten of the comments containing the monitored narratives, and no mass automated activity was recorded. These accounts were concentrated in specific channels, especially the Belarusian regime-aligned YouTube channel “Facets of Freedom” (Грани свободы). As election day approached, identical, repeatedly posted texts urged a boycott of the elections and a vote for parties that would restore relations with Russia, and on the day after the elections they disputed the results. However, these accounts played only a minor role in the discussion of the election outcome: under the main news items on the exit poll results they posted only about 6% of the comments.

Attitudes towards political forces were asymmetric. In the monitored channels, the government was almost always spoken of negatively. Sovereign Power (Suverēnā vara) and Latvia First (Latvija pirmajā vietā), meanwhile, were among the few political forces that received more positive than negative assessments. The data, however, do not reveal coordinated action between these parties’ accounts and the channels that amplified content favourable to them, although they do not rule it out either.

As election day approached, hateful and violent language was increasingly directed at politicians. By election day, more than 3,400 comments had been automatically classified as possible violations of platform rules, and the majority of them (62%) consisted of hateful or abusive language. Hatred was directed at Latvians and the Latvian state on the one hand, and at Russian-speaking residents on the other. Most of these comments probably reflected the users’ own attitudes, and inauthentic activity amplified them rather than created them.

Coordinated activity can be detected and curbed, but reducing the echo among the domestic audience depends on trust in state institutions, on timely explanation of decisions and on the quality of the discussion environment on the accounts of Latvian media and politicians. Recommendations for strengthening election integrity are set out at the end of the report.

About this report

This report was prepared by the Centre for East European Policy Studies (CEEPS). It is based on monitoring of Latvia’s information space that CEEPS carried out in cooperation with the social media analysis platform DisInt.ai from 1 July to 4 October 2026. The monitoring covered 502 public channels on YouTube, Telegram, TikTok, Facebook and X, and only publicly available data were used in the analysis. The research method, its limitations and ethical principles are described in Chapter 2 and in the Annex.

The conclusions expressed in this report reflect solely the views of the authors.