Fake Twitter Account Patterns – Their Achilles’ Heel

Share:

As mentioned previously, organized, agenda-driven, fake Twitter accounts are created in groups. Sometimes they are created in groups of a dozen or so interconnected accounts. More often, however, there are so many interconnected fakes that it’s impossible to find the bottom. Regardless of their size, the fake Twitter account groups have distinct, discernible patterns. And those patterns are their Achilles’ Heel.

The Fake Twitter Account Groups Got Busted

When I started tracking fake Twitter accounts, the female model bots followed each other. Find one model bot, go to its followers, and find hundreds or thousands of identical accounts, often with the same model on the profile. Catching on to this foundational pattern, a number of bot researchers called out the accounts, listing hundreds at a time for reporting (since Twitter’s algorithms seemed incapable of catching the pattern on their own).

The reporting started to have an impact. Twitter started suspending a large portion of those early-version model bots. It wasn’t long after this that I noticed a distinct change in the pattern of the model bot accounts. New, female model bots were no longer following each other, making it much more difficult for bot researchers to find the bot groups for reporting. It seems to avoid detection, whoever was creating and directing the model bot accounts had adjusted how they were distributing their fake accounts.

You Can Change the Patterns, But You Can’t Get Rid of Them

Just as with the aforementioned model bots, however, the new distribution channel has its own patterns. And, logically, this makes sense. Creating, building, directing, and tracking endless amounts of Twitter accounts must require high levels of organization.

Consider if you created just 100 fake accounts to boost your messaging (whether commercial or political). It would be too time intensive to create 100 individualized accounts. So, you might create 10 groups of 10 with the same profile picture and description. Then, you need to build your following for your messaging to be spread. So, you’d want your 100 accounts to follow one another to help build the accounts (which will help draw real followers).

But you’re smart! You don’t want every account to follow all of the others. This would be too easily detected by bot researchers. Instead, you might randomize your 100 accounts into 10 groups of 10 again and have each group follow just 10 of the other fake accounts. No matter which choices you make, patterns will manifest in your fake accounts.

The New Pattern

After the old model bots were reported into oblivion, a new pattern inevitably emerged. Instead of having the female model bots follow each other, the female model bots were now following fake male accounts and vice versa. Only a small percentage of the followers on these fake accounts are the same gender now.

Female Model Bots with Fake Male Followers

Fake Male Accounts with Female Model Bot Followers

Another Part of the Pattern — Interconnected Followers

Think it’s feasible that the male-female split in the fake accounts is some weird, cosmic happenstance? There’s more evidence that the pattern is purposeful.

Tracking the followers on these fake accounts, you find that the followers are almost identical, though not always in the same order. In this sample set, I’ve taken just the last nine followers from a small, sample set of the fake male accounts. The interconnected accounts are circled.

It’s Not Just Model Bots

If you track more than the last nine followers of these accounts, you’ll discover that American model bots aren’t the only interconnected fakes. This group of fake Twitter accounts also interconnects with (among others) foreign model bots, fake lottery winners, and fake celebrities.

Twitter’s Own Algorithms Know About the Patterns

“Twitter’s Own Algorithms Know About the Patterns” is quite a claim. But, it’s a claim that is easily proven. It’s been long-known that Twitter’s “recommended accounts to follow” algorithm is a tip-off for fake accounts within the same bot groups. The new female-male following pattern has not changed that.

The patterns become obvious when tracking these “recommended” accounts.

Let’s look at the three “recommended accounts” on this fake male account. Notice that they all share:
1 > Similar tweet numbers
2 > Similar Following and Follower numbers
3 > Almost all of the followers are female model bots

The first recommended account:

The second recommended account:

The third recommended account:

What Twitter’s “recommended” lists tell us:

Their algorithms know that these accounts are connected in some way. And this begs the question: Why doesn’t Twitter get rid of the obviously fake, interconnected accounts? They can’t claim that their algorithms don’t detect them.

Key Takeaways

There are several important items to note in the fake Twitter account pattern examined in this article.

1 > Those seemingly harmless fake lottery and celebrity impostor accounts are connected to this bottomless but interconnected group of fake accounts. As are the fake foreign model bots. This strongly suggests that seemingly harmless accounts might not be so harmless after all.

2 > The examples used in this article are a sliver of the thousands of these accounts I’ve documented. And the documented accounts are a sliver of the actual amount of these fake Twitter accounts. The magnitude of this one, specific fake account type is jaw-dropping.

3 > Model bots are just one type of fake Twitter account. But they are relatively easy to identify, which is why they make excellent examples. As “simple” as these accounts may seem, however, their patterns are indicative of how most fake Twitter account patterns manifest, including (but not limited to) bitcoin bots, political trolls, “follow-back” fakes, and commercial “like” bots.

4 > Twitter knows. And while Twitter executives continuously claim that they are cleaning up their platform—this article, in of itself, is a case-study in Twitter’s actual, minuscule efforts.

Written by Virginia Murr