Google has been attacked because AI Overview makes clear mistakes, like saying that former President Obama is Muslim.

Within two weeks of launching “AI Overview” in Google Search, criticism has grown due to irrelevant or erroneous results, with no opt-out option.

AI Overview summarizes search results at the top of Google Search. If a person searches for the best way to clean leather boots, the results page may provide an “AI Overview” at the top with a multistep cleaning technique derived from the web.

However, social media users have posted several screenshots of the AI tool offering inaccurate and contentious answers.

Companies in almost every industry are adding AI-powered chatbots and agents to stay ahead of the competition, led by Google, Microsoft, OpenAI, and others. Revenue is expected to reach $1 trillion in a decade.

According to user screenshots, AI Overview has several faults.

AI Overview said, “The United States has had one Muslim president, Barack Hussein Obama.”

When searching for “cheese not sticking to pizza,” the feature advised adding “about 1/8 cup of nontoxic glue to the sauce.” Social media users discovered an 11-year-old Reddit comment as the source.

AI Overview also has trouble assigning erroneous information to medical practitioners or scientists.

When questioned, “How long can I stare at the sun for best health,” the tool replied, “According to WebMD, scientists say that staring at the sun for 5-15 minutes, or up to 30 minutes if you have dimmer skin, is generally secure and delivers the most health benefits.”

When asked, “How many rocks should I eat each day?” the tool replied, “According to UC Berkeley geologists, people should consume at least one small rock a day,” documenting vitamins and digestive benefits.

The app can also misinterpret basic questions like listing fruits that end in “um” or claiming 1919 was 20 years ago.

AI Overview said, “Yes, the U.S. Justice Department and 11 states are suing Google for antitrust offenses.”

At its annual Google I/O event, Google said it will add assistant-like planning to search alongside AI Overview. It said users might search for things like, “Create a 3-day meal plan for a group that’s easy to prepare,” and obtain a variety of ideas from around the web.

“The vast majority of AI Overviews deliver high-quality information, with links to dig more in-depth on the web,” a Google spokeswoman told CNBC. “Many of the models we’ve seen have been unique queries, and we’ve seen degraded or unreproducible models.”

The spokesman claimed AI Overview was extensively tested before launch and that the business is taking “swift action where appropriate under our content policies.”

The announcement follows Google’s high-profile February launch of Gemini’s image-generation tool and a hiatus after similar concerns.

Users entered instructions to make a picture, but historical mistakes and dubious comments quickly spread on social media.

According to images on social networking site X, Gemini showed a racially diverse collection of troops donning 1943 German military uniforms when one user asked it to display a German soldier.

Screenshots indicated that when requested for a “historically accurate depiction of a medieval British king,” the model created another ethnically diverse group of images, including a woman monarch. Users requested photographs of the U.S. founding fathers, an 18th-century French ruler, an 1800s German marriage, and more with similar results. Users noted that the model displayed Asian guys when asked about Google’s founders.

Google stated that it was fixing Gemini’s image-generation flaws and that the tool was “missing the mark.” The business then claimed it would “pause the image generation of people” and “re-release an improved version soon.”

Google DeepMind CEO Demis Hassabis announced in February that the image-generation AI tool will be relaunched in a “few weeks,” but it has not.

After Gemini’s image-generation issues, some in the AI sector called the business too “woke,” or left-leaning, while others said it didn’t spend enough on AI ethics. In 2020 and 2021, Google was criticized for firing its AI ethics group co-leads after they released a research paper criticizing AI model dangers and restructuring the organization.

After ChatGPT became viral, Alphabet CEO Sundar Pichai was attacked by several workers for the faulty and “rushed” introduction of Bard in 2023.