Twitch is not unaware of the threats. A Twitch spokeswoman said the company planned in the coming months to livestream a session that will educate streamers about real-world risks. In recent years, it has increased its efforts to build safety into the platform, said Mr. Verrilli, the head of product. He noted, for example, a change the site made to obscure personal contact information on the Twitch settings page, so streamers sharing their computer screens wouldn’t accidentally expose their address or phone number.

Angela Hession, Twitch’s vice president of global trust and safety, said her team kept creators up to date on “how to protect themselves, both on Twitch and off,” including by offering a safety center with tips for preventing doxxing, swatting and stalking. Ms. Hession said Twitch tried to create “a safe environment” but was limited in how much it could do to help. It can’t, for example, give out identifying information about a potential harasser unless the company receives a valid request from law enforcement. The team at Twitch responsible for corresponding with law enforcement and informing it about threats made on the platform has quadrupled in the last two years.

Last year, the company announced it would begin holding users accountable for misbehavior that occurred “off-service,” saying it was a novel approach for the industry. If a Twitch user is determined to have committed “egregious real-world harm,” according to the company, the user can be barred from the platform.

Twitch has to walk a fine line between keeping streamers safe from unruly fans and encouraging the kind of interaction that powers the platform and makes money, said Mia Consalvo, a professor at Concordia University in Montreal who studies video games and Twitch.

“They want to shut down the most egregious harassment, because that’s going to drive people away from the stream and the channel, but they don’t want to crack down too much, because they don’t want to drive away too many people, too many viewers,” Dr. Consalvo said.

In 2020, Twitch expanded its definition of hateful conduct and acknowledged that some creators, especially minorities, “experience a disproportionate amount of harassment and abuse online.” Last summer, the hashtag #TwitchDoBetter began circulating on social media after Black and L.G.B.T.Q. streamers said they were being targeted by so-called hate raids, in which automated bot accounts spammed their chats with racist and discriminatory epithets.