Google Is Using AI to Cut Down on False Info in 'Featured Snippets' Results

pcmag.com:

Google's latest Search improvement aims to curb the spread of disinformation via its featured snippets sections—those text boxes that sometimes misrepresent the facts.

Previously, an algorithm pulled blocks of quoted text(Opens in a new window) from websites it determined answered users' questions, providing photos and a link to the source. But Google is now employing its Multitask Unified Model (MUM) AI to better understand what VP of Search Pandu Nayak calls(Opens in a new window) "the notion of consensus," or when multiple high-quality online resources agree on the same information.

By checking snippet callouts (the bold word(s) above the featured text) against various prominent sites, MUM can look for a general agreement—even if various sources use different words or concepts to describe the same thing. "We've found that this consensus-based technique has meaningfully improved the quality and helpfulness of featured snippet callouts," the company said.

AI models are also helping Google understand when featured snippets are not useful, particularly in cases that have no real answer. Take, for example, a search for(Opens in a new window) "When did Snoopy assassinate Abraham Lincoln?" While the site provides accurate information from Wikipedia about where, when, and how the president was murdered, it, as Nayak pointed out, "clearly isn't the most helpful way to display this result."

The new update reduced by 40% the triggering of featured snippets under false pretences, according to Google.

Google Search is also expanding "About this result," which launched last year to provide details about a link (i.e. a description of the page from Wikipedia) before you open it. Now, users can find even more context about that