Twitter says Musk’s spam analysis used tool that called his own account a bot

As explained many times by many people, the 5% is not the number of bots on Twitter. It is the false negative rate of Twitter's bot detection algorithms.

> No one is saying that 5% of Twitter's users is bots, except for when Musk is trying to put those words into Twitter's beak.

"In its disclosures, Twitter claims to have nearly 238 million monetizable daily active users (“mDAU”) who participate on the platform, and tells its investors that this userbase metric is a bellwether for its ability to generate revenue and the “best way to measure [Twitter’s] success . . . .”

I think they are talking about this number being significant less than 238 million.

"They show that in early July fully one-third of visible accounts may have been false or spam accounts—resulting in a conservative floor of at least twice as many false or spam accounts as the 5% that Twitter discloses for the entire mDAU population."

They are talking about the same after bot removal 5%.

Twitter makes what seems like a good "gotcha" point here, but knowing a little bit about how social media bots work, I'd like to try and give my 2 cents about why Musk's profile is probably not a great one for this type of analysis.

1) Many people who create a brand new Twitter account just end up following a few celebrities and then either not tweeting much or forgetting about the account after a short time after deciding that Twitter is not for them. It's likely that somebody well-known like Musk has a ton of actual human followers who would likely be labelled as bots because they're users with basically blank profiles that would be seen as bot-like.

2) When programming any reasonably sophisticated bot that doesn't just blurt out spam as fast as possible until its banned, the programmer would design it to blend in, to avoid detection by any defense mechanisms that Twitter would have. It would likely be seeded with an array of well-known Twitter profiles to follow at random to appear like a normal person with normal interests. Any particularly well-known Twitter account would be loaded with bots just through bot authors starting out by seeding it with a list of accounts they're familiar with.

3) Due to his life and wealth, Musk probably also exhibits characteristics that makes him seem more bot-like: such as probably tweeting in random bursts at odd hours in many different geographical locations.

4) Given his status as an international figure that also has a role in geopolitics and even military action, it's not impossible that there are state actors with big resources who create bots that follow Musks account for anything from trying to influence him with responses and through polls, or to try and make him look bad by "exposing" his bot activity at a later date when they need to influence public sentiment against him.

> First, it wouldn't matter anyway. He signed away any and all rights to due diligence in the agreement.

He waived DD but the claim is that Twitter specifically inflated numbers to make itself appealing for an acquisition/merger, which would still be grounds for breaking the contract without Musk needing to give any money to Twitter or complete the purchase.

No. Nobody outside of Twitter can repeat the analysis.

First, only Twitter knows which users are active (the population being analyzed are the DAUs). People doing bot analysis from publicly available define activity based on the account tweeting, which will probably skew heavily toward spam bots.

Second, the mDAU metric is the DAUs with known bots having been removed. Nobody outside of Twitter knows which active accounts were excluded by Twitter from the metric. Even if 50% of Twitter DAUs are bots, as long as Twitter detects 90% of them as bots and marks them as non-monetized, the 5% number stands.

Third, nobody outside of Twitter can actually do a proper job of evaluating whether an account is a bot, since they have orders of magnitude more signals than a simple tweet stream / public profile information.

Twitters methology is far better than any publicly available bot detection would be, but the flipside is that it's not a replicable methodology.