Users catch Google AI Overviews inventing fake facts

When you Google something, you expect facts, not fiction

But Google's AI feature might tell you to put glue on your pizza or explain the meaning of slap a goose

This is all thanks to AI Overview, a tool that's supposed to save you time, but might just leave you questioning reality

You may have noticed the new feature pop up above certain Google searches last May

According to Google, the feature will appear in Google search results when our systems determine that generative responses can be especially helpful

For example, when you want to quickly understand information from a range of sources, including information from across the web and Google's knowledge graph

But as social media users are pointing out, it seems to be making up definitions to random phrases or terms

Try a simple search of a random string of words and add the word meaning at the end

We tried milk the thunder meaning, which, according to AI Overview, is a metaphor that suggests using or exploiting a situation to one's advantage

But when you click the associated hyperlink claiming the source, the linked article mentions nothing about milking thunder and instead two separate phrases

The article called Weird English Phrases and Their Meaning by EF English Live lists steal someone's thunder and crying over spilt milk

Or one example from Wired highlighting a social media example, you can't lick a badger twice meaning, to which the AIO review says is an idiom meaning you can't trick or deceive someone a second time after they've been tricked once

If this all sounds familiar it because Google faced a similar issue during the Super Bowl Travel blogger Nate Haig pointed out errors in Google 50 short ads highlighting small businesses won from every state In Wisconsin

ad, fittingly set in America's Dairyland, Google's Gemini chatbot helped a cheese monger

write a product description, claiming Gouda accounts for 50-60% of the world's cheese

consumption. Haake fact-checked on X, saying Gemini provides no source, but that is just

unequivocally false. Cheddar and mozzarella would like a word. According to Google, an AI

hallucination is an incorrect or misleading result that an AI model can generate. Since models are

trained on data, they learn to make predictions based on patterns, but the accuracy depends on

the quality of that data. A Google exec replied to Haake saying, not a hallucination. Gemini is

grounded in the web and users can always check the results and references. In this case, multiple

sites across the web include the 50 to 60 percent stat, but they did quietly re-edit those ads

Many Americans remain skeptical of AI, and these latest glitches could fuel their doubts. Edelman's

2025 Trust Barometer study shows while 72 percent of people in China trust AI, only 32 percent of

Americans do. Edelman says some see AI as a force for progress, while others worry about its

unintended consequences. And according to Tech Times, Google spokesperson Megan Farnsworth says their system attempts

to offer context whenever it can, but nonsensical prompts are still likely to show up in AI overviews

With Straight Arrow News, I'm Kennedy Felton. Download our app or visit san.com for more