About 2,200 people watched a shooter video of their attack in front of a synagogue in Germany before being removed from the Twitch video streaming site.
Twitch, owned by retail giant Amazon, said five people saw the video when it was broadcast live.
The footage remained online 30 minutes after the live stream, during which more than 2,200 people saw it.
Twitch said the video was not advertised in its "recommended" feed.
"Our investigation suggests that people coordinate and share the video through other online news services," the company said in a statement.
The attack occurred in the city of Halle in East Germany on Wednesday around 1
The video shows a man making anti-Semitic comments in front of the camera before moving to a synagogue and shooting at her door.  After he did not get in, the shooter shot two people nearby.
The suspect is a 27-year-old German, who acted alone according to local media.
In a statement, Twitch stated that he has a "zero tolerance policy against hate behavior".
"Every act of violence is taken very seriously, we have been working hard to remove this content, and all accounts found to be posting or republishing the content of this vile will be permanently closed," Twitch said shared a "hash" of the video with a group of technology companies, including Microsoft and Facebook.
A video hash is essentially a "fingerprint" of a video that allows platforms to detect whether the same footage has been uploaded to their service
In March, an attack on a New Zealand mosque killing 51 people was broadcast live on Facebook
The social network was criticized for failing to prevent copies of Videos of the shootings in the Christchurch Mosque will be shared on its platform.
Facebook has since Plans discussed, training algorithms to detect videos of shootings so they can be detected and removed faster.
It is planned to use the material from police cameras captured during the training sessions to teach the real-life shooting video detection systems.
"We are still far from finding a solution to this problem," said Christopher Tegho, a machine learning engineer at video analytics company Calipsa.
"Understanding an entire scene is a harder and more complicated task.
" One of the problems is getting enough data to understand the shooting scenes. For this reason, Facebook asks the police to collect this data. This is the first step. "