Armenian diaspora Facebook groups mapped

TLDR

• DFRLab analysed 204,552 posts from 83 Armenian diaspora Facebook groups, identified via Meta Content Library and scraped with Apify.¹
• Twenty of 6,586 active accounts posted almost 20 percent of pre-election posts, mostly amplifying anti-government and pro-opposition sources.¹
• DFRLab also found 17 seemingly inauthentic accounts in two clusters promoting Gagik Tsarukyan's party.¹

DFRLab published a report on 4 November on political messaging in public Armenian diaspora Facebook groups before Armenia's 7 June 2026 parliamentary elections, using a data corpus built with Meta Content Library and Apify.¹ It was written with Qriton Technologies and Erebus, with data collection by Apify.¹

Why it matters: the report is a worked example of a method for studying a closed corner of Facebook: use Meta Content Library to find groups, a scraper to collect posts, a simple spike rule to flag narrative surges, and posting-time patterns to test for coordination. It also shows the method's limits, since DFRLab says its data could not show which posting tools one account used.¹

Driving the news:

• Corpus: 204,552 posts from 83 public groups, 1 March to 18 June, by 7,395 distinct authors. Two groups were covered only from later dates, and posts deleted before collection are not included.¹
• Concentration: the most active 1 percent of authors produced 38 percent of posts, and 65 percent of posts were reshares with little or no text. Twenty of 6,586 active accounts, under 0.4 percent, published 33,969 of 175,797 posts before election day.¹
• Spike rule: a topic spiked when its distinct posts on a day reached at least twice its median daily output over the preceding 14 days, so reposting alone did not register. In mid-March genocide vocabulary reached 2.9 times its baseline.¹
• Reach versus volume: the widest-reaching account posted 413 times across 38 groups, while the most prolific posted 1,283 times in 17 groups.¹
• Timing: one article appeared in at least 49 groups, 37 of them within 11 minutes. DFRLab says Facebook's built-in multi-group sharing could produce that speed but its data could not confirm which tool was used. Meta Content Library found copies outside the scraped groups, giving times to the minute where Apify gave exact times, and nine copies had no time because they were removed.¹
• Inauthentic network: DFRLab found 17 seemingly inauthentic accounts in two clusters, of 11 and six. On the evening of 4 May, ten accounts in the first cluster posted the same slogan, 15 copies into six groups between 20:40 and 21:13 CEST, with a new account joining every two to four minutes and most accounts using one group. The accounts had little personal information and some used profile photos taken from websites.¹
• Ethics: the report names accounts but not the people behind them. Accounts that appear to belong to real individuals carry a pseudonymous code derived from the platform identifier, while seemingly inauthentic accounts were not pseudonymised.¹

The big picture: Armenian citizens abroad could not vote from overseas, except diplomats, military personnel and their families, who could vote electronically. DFRLab says messaging in the groups could still aim to influence relatives at home, mobilise travel to vote, or deepen divisions between diaspora and government.¹ It reports that the share of diaspora-related posts carrying political messaging rose from 30 percent in early spring to 54 percent after the vote, and that most of the ten widest-reaching accounts criticised the government.¹

The other side: the sections reviewed carry no response from the party or from the accounts described.

Single-source note: everything here comes from the DFRLab report, which this article read in nearly full text but without the interactive charts; no independent replication was found, and a monitoring site only mirrors it. DFRLab states its own limits: group selection relied on group names, and the Europe and US set had more groups, so the widest-reaching accounts leaned toward it by construction. "Seemingly inauthentic" is DFRLab's assessment from posting patterns and profile checks, not proof of who ran the accounts. The report also mentions an identical photo posted by 11 accounts on 13 August 2026, after the dataset's 18 June end, which DFRLab says was found by manual check and is not in the scraped data.¹