A workaround approach - disregarding any existing standards - would be to simply give the user a random, unique string to search for and then try to detect that string in the URL (typically as a query parameter) - then the browser just has to replace that string with the user's actual search query in subsequent searches.
This is probably how Firefox works, because I get suggested a new search engine every time I search on a new site.