User-Agent: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.8.1.9) Gecko/20071025 Firefox/2.0.0.9 Build Identifier: Mozilla/5.0 (Windows; U; Windows NT 5.1; en-US; rv:1.9b1) Gecko/2007110904 Firefox/3.0b1 When only one or two characters are entered into the addressbar, autocomplete's first suggestions will be commonly visited pages with no immediately apparent relation to the characters entered. Reproducible: Always Steps to Reproduce: 1. Enter one or two characters into addressbar. 2. Observe autocomplete suggestions. Actual Results: The first autocomplete suggestions have no apparent relation to the characters entered, instead simply being very frequently visited sites. Representative examples: Entering "ne" will suggest the more frequently visited "http://forum.rpg.net/forumdisplay.php?f=29" instead of the less visited "http://www.newsarama.com" which is the second suggestion. Entering "w" will make this the first suggestion, but the second suggestion will now be "http://slashdot.org/" Entering "co" will first suggest "http://www.newsarama.com/" and "http://tvtropes.org/pmwiki/changes.php/" over any sites with URLs beginning with "co," such as "http://community.livejournal.com/scans_daily/" This is a regression from Firefox 2, where autocomplete properly suggested completions based on the entered characters even with only one or two characters. It seems, based on short observation, that autocomplete is considering any individual word contained in the title of the website as important or more important than the URL itself. If this is the case, a single frequently visited site with a long title might become the first suggestion for an undesirably large set of one to two characters beginnings. This is a regression from Firefox 2. Expected Results: Autocomplete should suggest completions with URLs starting with the entered characters (barring www. prefixes and the like) first. Entering "ne" should suggest "http://www.newsarama.com/" before "http://forum.rpg.net/forumdisplay.php?f=29", even if the later is a more frequently visited page, and so forth. Assuming my guess to the cause of the issue is correct, this might be fixed by "weighting" the words contained in the actual URL over words in the title for autocomplete.

you're right, the new places autocomplete weights results based on visit count rather than on url position of the found token.