Earlier this year, a brand started seeing unusual levels of engagement on Twitter. The account was generating more followers and more retweets. This would have been a welcome development — if only it weren’t mostly bots and fake accounts prompting the sudden popularity.
“We thought we were on to something. Like maybe we hit the sweet spot,” said an agency exec, who worked on the account and spoke on condition of anonymity. “This brand that organically wasn’t doing well, all of a sudden starts promoting content, and we noticed it was getting a ton of engagement.”
The advertiser had asked Twitter if any of the activity was from bots, which Twitter denied. So the agency decided to scrutinize the campaign itself.
It turned out that between 10% and 20% of Twitter activity on that particular campaign came from bots, according to the agency executive.
Bots are a sensitive subject for the service, which has fought to grow every legitimate regular. The official line at Twitter is that it now has 328 million monthly active users, and it doesn’t want its early reputation for hosting spammy bots to mar its future.
When, for example, Katy Perry just became the first celebrity to top 100 million followers recently, Twitter was quick to reject skeptics’ claims that many of those accounts were fake. It promised it verified the number.
And when Twitter opened to third-party measurement firms and let them audit its audience, it reported 99% human.
Still, when the advertiser that remained skeptical of Twitter’s bot counts dove into the campaign, it noticed many profiles were missing photos, missing followers, not tweeting, and other signs that they could be bots.
Brands remain concerned about bots, worried they might communicate with their accounts, follow them, retweet the content, and wind up costing money when advertisers pay for fake engagements. Brands make tempting targets for bots, because the brands often follow anyone that follows them, giving the bot some more touch points of legitimacy when they count top brands as contacts.
“Look at those accounts,” said Augustine Fou, a cybersecurity and ad fraud researcher. “When there’s no picture, no tweets, and they’re just retweeting. What would you call that if not a fake account.”
The bot problem is not new to Twitter, which estimated in 2014 that 5% of its accounts were bots. Twitter declined to comment for this story, but it has issued public statements regarding bots and its efforts to prevent their misuse. (Good bots can be helpful and share content that is welcome on the platform.)
Twitter has been dismissive of third-party statistics about bot activity, and claims that only it can measure that type of traffic accurately. However, Twitter does not disclose its methodology for determining bot activity, claiming that would give the bad actors more information to beat its system.
Pixelate, an outside measurement firm, said that Twitter has a bigger bot problem than its rivals. Twitter bots accounted for 3% to 17% of traffic coming from Twitter to outside websites measured by Pixelate in June.
Meanwhile, Facebook saw up to 3% of its traffic coming from bots, Instagram saw 4% and Pinterest saw between 2% and 9%, according to Pixelate’s analysis form June.
Twitter has opened to third-party measurement firms Integral Ad Science and Moat, and has said that the problem of bots is minuscule when it measures video views. “Among the video impressions Moat has measured on Twitter, Moat has detected that over 99% of them are viewed by humans,” Twitter reported when it first announced its new partnership.
Still, brands and agencies are skeptical of metrics being shared by all platforms, not just Twitter. Video is becoming a more ubiquitous format, and audience numbers will need to be trusted in order for brands to move their dollars to the social platforms.
Twitter runs video ads amid content from publishers like Bloomberg, BuzzFeed, Mashable, NBC and a number of others. Last year, presidential debates and National Football League games got more than 3 million people click into some of the video streams.
“Brands and agencies aren’t dumb,” said another agency exec, speaking on condition of anonymity. “We simply want a reliable way to measure what is real. They need to create a reliable system to detect the difference between humans and bots, because no brand wants to pay for that impression or engagement.”