Google’s AI-generated search results are already being slammed as a “disaster” that “can no longer be trusted,” the New York Post reports. Google shot back Friday to say it is taking “swift action” to improve AI summaries of search results.

Absurd “AI Overview” glitches began right as Google began rolling out the feature to U.S. users last week, with plans to offer it to 1 billion users by the end of 2024.

Its bizarre recommendations include adding 1/8 of a cup of non-toxic glue to tomato sauce to help cheese stick to pizza, along with the health benefits of tobacco for kids.

The glue-on-pizza advice was later traced back to a child's Reddit post, leading many critics of AI to wonder if it believes every piece of information it consumes.

“Tobacco contains nicotine, which can cause some short-term benefits such as increased alertness, euphoria and relaxation,” AI Overview responded to a query on “health benefits for tobacco for tweens.”

It also falsely claimed that tobacco can be used for “whitening teeth.”

When Google search is asked which U.S. presidents went to the University of Wisconsin-Madison, AI Overview claims that President Andrew Johnson earned 14 degrees from the school, the most recent in 2012. Johnson died in 1875.

AI Overview also falsely said Barack Obama was the first Muslim U.S. president.

“Google’s AI search results are a disaster,” says Tom Warren, senior editor of The Verge high-tech blog. “I hate that it’s turning into a resource that can no longer be trusted.”

A Google spokesperson responded, “The examples we’ve seen are generally very uncommon queries, and aren’t representative of most people’s experiences. The vast majority of AI Overviews provide high-quality information, with links to dig deeper on the web.”

Reviewers, however, say AI Overview demotes links to other websites.

Google users took to social media to lampoon erroneous AI Overview responses to queries such as whether people should eat stones or stare into the sun.

Users sharing their AI Overview experiences on X cited many other bizarre results. One person asked Google AI “how to treat appendix pain at home” and was told to try boiled mint leaves and a high-fiber diet.

Another user, apparently trying to trip up AI Overview, searched for “1000 km to oranges,” mixing distance with fruit.

Google’s response? “One solution to the problem of transporting oranges 1,000 kilometers is to feed a horse 2,000 oranges at a time, one for each kilometer traveled, and sell the remaining 1,000 oranges at the market.”

Some of the Google AI searches users denounced “were doctored, or we couldn't reproduce,” the Google spokesperson continued. “We appreciate the feedback, and are using these examples to develop broader improvements to our systems.”

Artificial intelligence critics fear that AI, especially from a source as trusted as Google, could rapidly spread disinformation without proper protections in place.

Google runs a disclaimer on all of its AI Overview posts stating, “Generative AI is experimental.”

Google is under growing pressure from AI-powered search engines like Perplexity, and from the repeated rumors that OpenAI, the creator of ChatGPT, is building its own AI search tool.

Searches through AI chats have also appeared on Facebook, Instagram and WhatsApp, with users able to get information from the web without Google.

These alternatives are praised by some for their cleaner experience than the often-cluttered results of a classic query.

Clearly, even though Google is now in its 26th year of business and its search engine is ubiquitous, its AI Overview is not yet ready for prime time.

With reporting by AFP