A web application might load the data behind a link as soon as the link scrolls into view or the pointer rests on it. Frameworks such as Next.js do a version of this for internal links, and browsers offer built-in hints for it. The same idea appears lower in the stack, where processors and databases read ahead because the next request is predictable.
The trade-off is wasted work. Anything fetched and never used still costs bandwidth, server time and sometimes money, and prefetching a page with side effects or personal data can cause real problems. It suits read-only pages with a likely next step, and it should stay away from anything that changes a record when loaded.