Twitch API code reveals plans for a 'Brand Safety Score' for streamers

This article has been updated to include a statement from Twitch.

Cybersecurity researcher Daylam Tayari has found code within Twitch's internal API that suggests that the streaming site plans to define "Brand Safety Scores" for individual streamers. The score appears to grade how brand-friendly a streamer is, using variables such as the age of the streamer (whether they're 18 or older, or 21 or older), their ban history, their relationship with Twitch, how the channel implements Automod, the ESRB rating of the game being streamed, whether the channel is set to mature audiences, as well as a manual rating and keywords for the channel set "by a Twitch-affiliated reviewer."

In the code, the score is referenced in relation to ads. A comment reads: "Grabs the Brand Safety Score of a channel as well as relevant data used to calculate it. Also returns custom parameters about this channel to forward to VAES for ad targetting purposes."

In a statement sent to PC Gamer, Twitch acknowledged that it is working on its ad targeting system, but said that it hasn't launched anything yet.

"We are exploring ways to improve the experience on Twitch for viewers and creators, including efforts to better match the appropriate ads to the right communities," said a Twitch spokesperson. "User privacy is critical on Twitch, and, as we refine this process, we will not pursue plans that compromise that priority. Nothing has launched yet, no personal information was shared, and we will keep our community informed of any updates along the way."

Twitch has added an automatic Brand Safety Score which grades how brand friendly every streamer is based on things like chat behavior, ban history, manual ratings by Twitch staff, games played, age, automod and more (See below).1/5 pic.twitter.com/VBl4HjGv7tMarch 9, 2021

Allowing advertisers to avoid having their ads appearing alongside content they find objectionable is fairly standard practice on the web. In theory, this allows advertisers to ensure their ads find an appropriate audience. In past practice, on platforms such as YouTube, this has sometimes limited the income potential of video creators who produce content for mature audiences, or who discuss news topics considered controversial. That in mind, some streamers have found the code worrisome. 

Since Tayari's posts on Twitter, there's been a great deal of speculation on social media about what the Brand Safety Score will or won't be used for. One obvious theory is that it could be used with Twitch's Bounty Board, a program that lets streamers automate the process of brand deals by playing games or watching videos with their audience for a payout. There are also scenarios where certain ad targeting variables are obviously important: For example, in the United States, an alcohol brand would only advertise on the channel of a streamer who is 21 or older. 

For some streamers, though, the idea that a hidden ratings system that could affect their opportunity to earn is a scary one. According to Twitch's comment, though, nothing has happened yet, and we should expect some kind of communication from the company if something does—it often posts updates on its blog.