Google is testing a new AI feature in its search results. While it's designed to make it easier to find information, it sometimes shows surprising results. Internet users are using it to trick it... with completely invented sentences.
Presented at last year's Google I/O conference, the AI Overview feature introduces a summary generated by artificial intelligence at the very top of search results. The goal is simple: to display a summary of relevant information without the user having to click on links. The system, powered by the Gemini model, is capable of analyzing web pages to derive a direct answer. Since the end of 2024, Google has extended this functionality to more than 100 countries in six languages, including English and Portuguese. But France, like other European countries, remains to away from this deployment.
For now, it is therefore impossible to access it without a VPN. Curious French users must go through another location to test this AI in the results. And those who tried quickly discovered an unexpected use: typing completely invented expressions, as if they were real proverbs. Result? Artificial intelligence does not detect the absurdity of the request and generates, with aplomb, a explanation and a plausible origin.
Google's AI summaries invent proverbs and find convincing meanings for them
One example that went viral was the completely invented phrase "A duck dog never blinks twice." Upon typing this phrase into Google, the AI claimed it was a proverb describing a hound so focused that it wouldn't blink. Upon relaunching the search, another explanation emerged: this time, the expression referred to a hybrid animal, half duck, half dog, used to illustrate something so absurd that it is hard to believe. Each new search thus produces a different interpretation… always false, but confidently told.
This behavior is as amusing as it is worrying. Because while AI may seem harmless in a humorous context, it remains a source of information visible to the user. Ultimately, when it is deployed in France, it could become a viral game... but also a new source of involuntary disinformation. A simple invented proverb can today become a "truth" interpreted by a machine. Further proof that artificial intelligence, however advanced it may be, still needs a little common sense.
0 Comments