14 DAY TRIAL // Google is introducing an updated Autocomplete feature which can now predict a search term even if it hasn't been typed ever before. Normally, Autocomplete relies on previous searches to determine what a user is typing. But if it's a new search, like most queries are each day, this doesn't work.
Now though, the algorithm looks at part of the query as well when making predictions, even single words, so it can provide a suggestion in a lot more cases.
"Today we’re improving the predictive powers of autocomplete, helping you search for things even when no one else in the world has," Bartlomiej Niechwiej, software engineer at Google, wrote.
"One of the main ways autocomplete works is by looking at the most popular searches on Google. For example, most people who type 'w-e-a' are going to search for [weather], so Google can make that prediction," he explained.
This works great for popular queries and Google does see quite a lot of similar or identical searches. Everyone has something they search for on a regular basis and some queries are just going to occur quite often.
"What’s tricky is that a huge percentage of the queries we get have almost never been typed before, so this makes it difficult to provide predictions based on popularity," he said.
"Now what we’re doing is making predictions based on only part of your search—specifically, the last word or words," he announced.
This refinement makes a lot of sense for unique queries or longer, more complex searches. While no one may have typed the exact same query, portions of it may have. At the very least, some words are going to be common enough that Google will know what you're trying to type.
Even in these limited occasions, Autocomplete is going to save you time. The feature is live on Google.com using English, but it will be rolled out to localized versions soon.