Google AI search tells users to glue pizza and eat rocks

Google began trialling AI overviews in search results for a small number of logged-in UK users in April, but launched the feature to all US users at its annual developer showcase in mid-May.

It works by using AI to provide a summary of search results, so users do not have to scroll through a long list of websites to find the information they are seeking.

It is billed as a product that “can take the legwork out of searching”, external though users are warned it is experimental.

However, it is likely to be widely used – and trusted – because Google search remains the go-to search engine for many.

According to web traffic tracker, Statcounter, external, Google’s search engine accounts for more than 90% of the global market.

It is still fundamental to the way in which Google makes its money, and a service the firm needs to both protect and future-proof.

Many industry experts agree that more focused AI-driven search is the way forward – despite the power-hungry tech’s environmental price tag.

Why wade through pages of search engine results and adverts to find information if a chatbot can give you a single, definitive answer?

But this only works if you can trust it.

So-called hallucinations by generative AI tools are not just a problem for Google, but as the world’s largest search engine it gets more scrutiny.

In one baffling example, a reporter Googling whether they could use gasoline to cook spaghetti faster was told, external “no… but you can use gasoline to make a spicy spaghetti dish” and given a recipe.

We don’t know how many searches it got right (because they’re less funny to share on social media), but AI search clearly needs to be able to handle anything thrown at it, including the more leftfield.

Rival firms are facing a similar backlash over their attempts to cram more AI tools into their consumer-facing products.

The UK’s data watchdog is looking into Microsoft after it announced a feature coming to its new range of AI-focused PCs that would take continuous screenshots of their online activity.

And ChatGPT-maker OpenAI was called out by Hollywood actress Scarlett Johansson for using a voice likened to her own, saying she turned down its request to voice the popular chatbot.