It's a request the front-end developer should be confronted with, not OpenAI.

The website could as well buffer the incoming stream until the used clicks an area to request the display of the next block of the response, once he has finished reading the initial sentences.