In Turkey, where twenty years of Recep Tayyip Erdoğan’s rule has seen the country descend into authoritarianism, mainstream media is almost completely under Erdoğan’s control. Social media remains one of the only outlets for those opposing Erdoğan to voice their discontent. Since the 2013 Gezi protests, the AKP has been trying to control the narratives on social media by employing social media trolls: networks of fake accounts that disseminate propaganda for the AKP and targets the opposition. Studies show that these troll networks are becoming more sophisticated and are changing tactics. Fazıl Alp Akiş explains how.
It was during the days of the Gezi Protests in 2013, when hundreds of thousands of Turkish citizens took to the street to protest then Prime Minister Erdoğan’s policies, that the ruling Justice and Development Party (AKP) realised the power of social media. Although Erdoğan’s AKP had coerced, co-opted, bought or otherwise taken control of much of the mainstream media, the Gezi Protests that took place in the summer of 2013 showed that the multitudes on social media were able to have their voices heard and build a strong political narrative. Prompted by the realisation that their control of the media falls short on digital platforms, Erdoğan’s advisors started looking for ways to have a presence on social media platforms.
Emails hacked and shared by Red Hack show that merely weeks after the peak of the protests, an advisor told Erdoğan’s son-in-law Berat Albayrak, Energy Minister at that time and former Minister of Treasury and Finance, to “set up a team of professional graphic designers, coders and former army officials who had received training in psychological warfare.” In three days, a feasibility study was conducted regarding “how the youth can be influenced by humour and slang while delivering a message and how opposition media organisations can be undermined by attacking their employees.” Thus, as a reaction to the Gezi Protests, the AKP created its first “troll army” to counter critical narratives about the government and to undermine the opposition. Inspired by the name of the ruling party, which organised the internet trolls for its propaganda business, the troll force has come to be known in Turkish politics as the “Ak-trolls”.
Trolls: relevance, modus operandi and function
Kemal Kılıçdaroğlu, the leader of the main opposition party CHP, announced in January 2022, in an address to the public, that he had commissioned a study of government-controlled Ak-trolls and claimed to have found a “gigantic network of trolls being paid by the treasury”. This brought the presence of Ak-trolls, known about for almost a decade now, back into the public debate.
What are trolls?
In internet culture, “trolls” refer to online accounts that interact with other accounts in a manner that distracts from the conversation and provokes an emotional response such as frustration. However, in the Turkish context the word “troll” is used as an umbrella term for any social media account that is perceived to be spreading propaganda, is most likely paid and is anonymous.
Under the umbrella of “trolls” we see that there are accounts that share automated and computer-generated posts, and accounts that are controlled by actual people that act in coordination with one another and are employed for propaganda purposes. Internet researchers call the former group “bot accounts” and the latter group “astroturfers”. In Turkey, the trolls are most active on Twitter and are doing Erdoğan’s and the AKP’s bidding.
The main objectives and methods of trolls and their employers can be classified as: initiating propaganda against targeted individuals and groups; hashtag campaigns aiming to create a new narrative, distract from another issue or magnify support in favour of a political group; working in coordination to “spam” other users through the online platform’s mechanisms so as to get the targeted accounts banned from the platform; spreading false information; and leaking documents that could not be shared in the mainstream media.
A history of Ak-trolls
Leaked by the activist group Red Hack, Albayrak’s emails show the birth of Ak-trolls, and later reports and studies enable us to track how the group enlarged and adopted new methods to increase their online influence. As the public’s awareness around the trolls increased, platforms increased their safeguards against malicious troll activity. However, troll networks adapted and changed accordingly.
The initial group of Ak-trolls were, for the most part, bot accounts active on Twitter that shared computer-generated posts alongside a hashtag in favour of Erdoğan and his government's policies. These accounts would create tweets using the hashtag and also interact/retweet with tweets using the same hashtags. A 2016 study published by the cyber security company Norton Symantec shows that among countries in Europe, the Middle East and Africa, Turkey is the country with the most bot accounts on Twitter. However, as bot accounts are easily recognisable and prone to suspension from the platform, they were replaced by more advanced troll networks.
Since that first mass use of Ak-trolls, the trolls have changed their network styles and methods quite a bit. The government-controlled trolls have become the subject of many academic studies, newspaper articles and political discussions. In 2020, Twitter announced that it was suspending 7,340 fake accounts that had shared over 37 million tweets from its platform. In the name of transparency, Twitter shared information regarding the accounts with Stanford Internet Observatory (SIO) prior to the announcement of the takedown of these accounts. Both Twitter and the SIO team (that I was a part of) found that the suspended fake accounts were centrally linked to the AKP’s youth branch. The takedown and the report published by the SIO made headlines in Turkey and opened up the presence of trolls to public discussion. (My colleagues and I were branded as American agents with terrorist sympathies for our work analysing the takedown.)
In our analysis of Twitter’s takedown, we found that the Ak-trolls expanded beyond bot accounts. Astroturfer accounts were organised in the form of “retweet rings” where the accounts formed groups with the sole purpose of retweeting each other. These groups were easily identifiable with names such as “Ak Davam Retweet Groups 1&2”. The accounts followed by the Retweet group would “retweet the same content, engage in inauthentic audience building tactics and work to falsely popularise hashtags”.
These retweet rings allowed the Ak-troll network to inauthentically popularise hashtags with a relatively small number of accounts with few followers. The retweet rings show us that the Ak-troll network has become more sophisticated, both in terms of accounts moving from bots to astroturfer accounts and their coordination with one another.
This latest announcement by the opposition leader Kılıçdaoğlu indicates that since the 2020 Twitter takedown, Ak-trolls might have moved into yet another level of troll activity: influencer trolls. These are not an alternative to bot or astroturfer accounts but a booster, as these accounts have a very high number of followers (both fake and real) and act as opinion leaders. In his address, Kılıçdaoğlu exposed a number of these influencer-trolls that specifically target him.
Spotting trolls is not especially difficult: they tend to not have real pictures or a real name associated with their account, and their profile activity in terms of tweets, retweets, likes and replies revolve around similar political issues. In most cases, these accounts would not have original or non-political tweets.
Spotting troll networks is even easier: the accounts tend have the same creation date, similar naming patterns (a name followed by four random numbers), similar bio descriptions on their accounts, their tweets follow the same flow (a support message for the employer political group, a picture of the supported political figure, or use of the same hashtag) and they often have few followers. These accounts are centralised, namely that they work in a coordinated manner prompted by central command, so they are usually active around the same time and tweet almost simultaneously.
Troll activity on Turkish-speaking Twitter
With influencer-trolls, astroturfers, bots and retweet rings, government-controlled trolls have a strong presence on Turkish-speaking Twitter. These networks of accounts work in concert to spread propaganda and misinformation. Detailed in the 2020 SIO report, troll activity peaks either in preparation for a ballot or to counter an opposition narrative. For example, when a mass online movement asked for Erdoğan’s resignation using the hashtag #TAMAM (meaning “enough”), bot activity peaked to get another hashtag trending: #DEVAM (meaning “onwards”).
A large part of their roles is getting the right hashtag to trend on command. These hashtags are sometimes in support of Erdoğan and his policies, and at other times against the opposition; they are most easily recognised when aiming to counter an organic hashtag created by the opposition. By these hashtag activities, the Ak-trolls aim to inflate their popularity and attack their opponents. However, this is not their most sinister use.
Since its birth, the Ak-troll operation has served to target public figures, most often journalists. Turkey is the world's second biggest jailer of journalists after China. Journalists are jailed on all kinds of charges, from terrorism to espionage, and with almost all mainstream media owned and controlled by Erdoğan, social media and online outlets are the last frontier for independent journalism. It is, therefore, on these platforms that the government targets independent journalists. Often, either just before or just after a government official points a finger at a particular journalist, the Ak-trolls start harassing the journalist by replying to his or her tweets with accusations and reporting the account in the hope of getting it suspended. By repeating these accusations, the trolls aim to damage the reporter’s reputation and implant the accusation in the public debate. Sadly, there have been more than a few cases where the targeted journalists were jailed as a result (or with the help) of these targeting campaigns, most recently Sedef Kabaş.
As reports that pointed attention to the AKP’s troll army came out and the opposition brought the matter to public attention, the government needed to respond. The initial reaction was denial: to the 2020 SIO report that exposed the existence of a troll army linked to the AKP’s youth branch, the AKP responded by saying that Twitter censored the speech of patriotic Turks and the people behind this operation were sympathetic to terrorists. Then, the government, run by presidential decree, made it mandatory for social media companies to have offices in Turkey and for the platforms to abide by Turkey’s censorship regulations. The platforms objected at first, but as the initial deadlines passed and the companies suffered millions of dollars in fines, they gave in.
The AKP wanted to further reverse the situation in its own favour, so their coalition partners opened up a debate on anonymity and presented a proposal that would mandate platforms to ask for proof of identity upon registration. At a superficial level, many may see anonymity as a problem partly causing trolls; however, anonymity is also a shield for freedom of expression in countries where political targeting of dissident speech occurs on a daily basis.
Yet another interesting approach on behalf of the government came when a top AKP official said that there were no trolls linked to the AKP, but they did have 2.2 million party members present on social media. This begs the question: where is the line between activism and trolling? If party members would like to share a common message on social media, is that necessarily a problem for democracy? Well, we can think about this problem as we do about physical gatherings: political gatherings happen all the time but when people show up to these gatherings with the promise of a paycheck, the act of gathering loses its democratic character.
It is clear that the problem of trolls, namely agents that spread false information, target people, and disseminate propaganda on behalf of their employer, is not a novel issue.. Such agents have always existed and now we see how they organise on social media. Although these agents put on a digital mask, my observation is that the government has reason to fear the people behind that mask. I would not be surprised if these agents, the ones behind the mask who see that the current regime is on its way out, start to speak about this operation in detail in the coming months.