“Eat at least one small stone daily.”

Google's new AI overview is no longer aimed at seriously answering silly questions – for example those about how many stones you can eat each day. The program has been modified to take into account less sarcasm and jokes, search engine head Liz Read wrote in a blog post. We should also use less user-generated content, such as forum posts, if they may lead to misleading advice.

Detailed answer instead of just web links

The overviews created with the help of AI aim to give users a more detailed answer to many search queries rather than just a series of web links. Google rolled out this feature widely in the United States in May. But shortly thereafter, embarrassing and even annoying errors in the program spread online. So in one of the “AI Insights” it was said that scientists advised to eat at least one small stone per day – and the source was a satirical article from the satirical site “The Onion”. Based on a joke on the online Reddit platform, it has also been recommended that the cheese be glued to the pizza using “non-toxic glue.”

Sarcasm as an answer

In the blog post, Google attributed this to, among other things, the fact that there are no really serious answers to some unusual search queries like “How many stones should I eat per day?” The program has filled this knowledge gap by providing access to satire. Google tested this functionality extensively before rolling it out widely — but nothing compares to millions of users trying out new search queries. Meanwhile, some alleged AI overviews circulating online were fake.

In other countries by the end of the year

It is expected that this functionality will be rolled out to other countries by the end of the year. Many website and media owners worry that Google's AI summaries will attract fewer people to them and their business will suffer. Google counters that there is actually more traffic to information sources that end up in overviews. How things go with the rest has remained unclear so far.