Google has announced some ways that it’s attempting to improve search results. The most interesting of these is the use of an updated AI model called Multitask Unified Model (MUM). With the latest update, Google has incorporated consensus into its results so when multiple sources agree on a fact, that’s what will be presented at the top of your results when asking questions.
With the new update, Google’s systems can identify snippets, even if different words are used, from high-quality sources to find out whether there is agreement on answers to queries. In its internal testing, Google said that the consensus method produced better quality answers to queries.
The new update also addresses questions where there are no answers. Google said if you type “when did Snoopy assassinate Abraham Lincoln” into search, you’ll be shown the accurate date and other information about Lincoln’s assassination, but obviously, Snoopy wasn’t the assassin, so it’s not a very good result. Google’s systems now detect peculiar cases like this and has reduced the triggering of featured snippets by 40%.
Aside from the better answer snippets, Google is expanding advisories where there are information gaps. According to the company, when there is breaking news, it’s possible for facts to get lost. In these situations, Google doesn’t want users reading something in its search results which turns out to be untrue.
In these cases, Google will show advisories at the top of its results to let you know that a topic is still unfolding and that it may be better to check results later when more facts are gathered. The advisory also warns people to check the source and to try new search terms.