Twitch CEO: AI Monitors for Brand Safety

Twitch CEO Dan Clancy shared new insight into Twitch's AI listening system on a recent episode of the Snarketing Podcast. When asked by the hosts about viewbotting and brand safety, Clancy shared the following:

"We also lately have been using AI to also identify streamers that might not match a brand because we transcribe streams to figure out oh is this appropriate for this brand and then also to determine if something is happening that may not match a brand. So that's something that we've had to invest a lot in to make sure that for a brand their content is appearing next to creators that and content that fits what they're trying to achieve."

Context:

Twitch's AI is always listening. You cannot opt out. Earlier this year, Twitch AI started listening to every stream at all times. This has enabled features like Twitch Clip That, Auto Clips, Sponsorship Campaign Assist, Stream Summary, the upcoming Stream Coach, etc. It's been assumed that Twitch was using the AI for brand safety purposes, but it was never confirmed - until now.

My Thoughts:

I think it was somewhat inevitable that Twitch would use their always-listening AI to apply brand safety standards to the advertising product. They tried this 5 years ago, with manual reviews of channels and a Brand Safety Score, but having an AI do it obviously lightens the human effort required. That said, I'm guessing the direct impact on ad revenue is small for now, but as they gain confidence in the system, they'll really tune it up. If/when that happens, brand-unsafe creators/content will see a consistent decline in ad revenue until they become safer.

Check out the conversation about this on Twitter, Bluesky, or Threads.