“I used to spend the whole day in front of the computer, starting early in the morning,” says 21-year-old Pedro from the city of Vitoria, in south-eastern Brazil.
“I posted pictures, wrote about my days, added people. And then I would give my opinion about some politicians, especially when there were debates between candidates going on on TV.”
It might sound like an average day for an ordinary young social media user, but Pedro (not his real name) is actually describing his time as a “cyborg”, someone who is paid to run fake social media accounts to influence public opinion.
Three years ago, during Brazil’s hotly contested general election campaign, Pedro says he worked for a Rio-based PR company whose clients include a number of leading politicians.
He says that for around $360 (£270) per month, he ran 20 fake accounts on Facebook and Twitter designed to create a buzz around the company’s clients.
BBC Brasil spoke to Pedro as part of an investigation into the way fake social media accounts were used during the country’s highly charged 2014 general election.
Read the full investigation (in Portuguese): Investigação revela exército de perfis falsos usados para influenciar eleições no Brasil
There is no evidence the accounts in any way influenced the result, or even that any of the candidates were aware of what was happening, but the investigation offers a fascinating glimpse into a new frontline in the way politics are conducted and elections are fought in Brazil.
‘Everything was monitored’
Pedro’s story is backed up by the accounts of three other young people who also worked as social media “activators” during the 2014 campaign.
- Facebook uncovers ‘Russian-funded’ misinformation campaign
- ‘You are being programmed’
They all told BBC Brasil that they worked from home via Skype.
“Everything was monitored. If I was online and didn’t respond, I could be penalised. So I had to inform a coordinator every time I took a toilet break.”
At the start of the job they were each given a set of fake profiles and photographs, plus basic personal details.
Their first task was to spend a couple of months building up or “activating” the profiles, posting everyday stories to establish themselves as real people.
After a while the activators would start talking politics.
And gradually they would begin interacting with each other, and then with real people. building up a network of friends.
The activators often used the social media management platform Hootsuite, to control many accounts simultaneously.
They would praise whichever political candidates they were being paid to support, attack their opponents and sometimes join forces with other fake accounts to create trending topics.
“Either we would win [debates] through sheer volume, because we were posting so much more than the general public could counter-argue,” one activator told BBC Brasil. “Or we would manage to encourage real people – real activists to fight our fight for us.”
Dead woman’s photo
The inclusion of personal details and non-political posts made the accounts much harder to detect because they broke up the pattern of automation which usually helps social media platforms spot fake accounts.
Operators of such accounts are often called “cyborgs” because of this mix of automated and “human” posts.
BBC Brasil’s investigation identified at least 100 fake accounts on Twitter and Facebook which were active during the 2014 election.
They all used photographs sourced from image stocks or stolen from news websites and existing social media profiles.
BBC Brasil was able to trace several of them. One turned out to be a female murder victim whose photograph had appeared in a local newspaper. Another was a well-known actor from Greece.
Some images were digitally modified to make them hard to track. This was the case with a picture stolen from Rio journalist André Moragas, and used by the fake profile of “Jonh Azevedo”.
Jonh’s account was created in 2012 and for a couple of years just posted personal messages.
“Son finishing another semester at university!” said one. “Very proud!”
During the 2014 presidential elections the “Jonh Azevedo” account suddenly went political, posting in support of an opposition candidate.
“I suppose they plant a fake profile and leave it to mature,” says Andre Moragas, whose photo was used by “Jonh”.
“This guy didn’t appear yesterday, but five years ago, connecting with people, gathering followers.”
Eventually Jonh Azevedo was debunked as a fake after other Twitter users became suspicious of his repetitive style. They reported him after the account posted the same phrase – “Need to rest” – twenty times in two months.
Some profiles appear to have been “recycled”.
BBC Brasil found the same user, “Fernanda Lucci”, appearing in three different conversation threads in three different states in 2014 appearing to support three different candidates.
Looking back, the four activators have mixed feelings about the job they were doing.
“I was a bit naive at the time,” one woman told BBC Brasil. “I had limited access to the accounts and couldn’t check them.”
But another has no regrets.
“You are just a person masking behind a profile,” he said. “The response is strong, the interaction is good. You feel that you really make a difference in a campaign.”
Many of the fake accounts identified by BBC Brasil have been inactive since the 2014 election.
Both Twitter and Facebook have told the BBC they are constantly working to identify, suspend and remove fake accounts.
That task looks set to be even more challenging next year as Brazilians prepare to go to the polls again.
Facebook deleted tens of thousands of fake accounts in France and Germany ahead of elections in both countries this year, and the organization told BBC Brasil similar measures were now being considered in Brazil.
The 2018 campaign is expected to be even more bitterly fought than four years ago and, with “fake news” and “troll factories” now common currency around the world, everyone will be keeping a close eye on the role of social media.