If you’ve ever searched for something on Google but came across a small info box highlighting the top answer, you’ve come across one of Google’s favorite snippets. Featured snippets are small Google results that the search engine packs up and delivers to the top of the page for many search queries.
The problem with featured snippets is that from the user’s point of view, these results seem very reliable – after all, they appear at the top of the results page. Since Google first introduced them many years ago, they have become more common over time, but like the rest of Google search results, snippets are algorithmically populated, not programmed by human curators.
Google says it’s implementing a stealth change that should improve the answers people see in these information boxes at the top of many search results pages. A new AI model called Unified Multitasking Model allows the search ranking system to validate its own work, according to Google. The AI model achieves this by matching the top bold text part of the search result against established high-quality search results to see if they say the same thing, even if they do so with different wording.
“We found that this consensus-based method significantly improved the quality and usefulness of featured snippet captions,” Pandu Nayak, VP of Google Search, wrote in a blog post.
Another problem, according to Google, is that sometimes a search engine will give reasonable answers to a search query that is itself erroneous. Google’s latest AI model should also help its result ranking system understand when displaying results in a snippet is inappropriate because the premise of the question is wrong. The company says featured snippets now appear 40% less on these occasions.
“This is especially useful for unanswered questions: for example, a recent search for ‘when Snoopy killed Abraham Lincoln’ returned a snippet with a specific date and information about Lincoln’s assassination, but this is clearly not the most useful way to show this result,” Nyack wrote. .
Google also announced that it will expand the use of warning messages for searches that do not produce results that the search engine has “high confidence in”. The company is already using these content guidelines for new topics that lack established search results, but says it will now use them when the overall search results do not meet its quality standards.