‘There are two ‘P’s in the word Google’ says the company’s upgraded AI Overview, as an old LLM issue rears its ugly head
Google’s recently upgraded AI Overview feature, which provides conversational, AI-generated responses atop search results, has demonstrated a fundamental limitation of large language models: the inability to accurately handle basic factual information. The feature incorrectly claimed there are two letter ‘P’s in the word “Google,” which contains zero P’s—a basic spelling error that highlights an old and persistent issue in LLM technology.
The AI Overview represents Google’s broader push to integrate more AI capabilities throughout its products. While the upgrade aims to provide more natural, conversational answers to user queries, the apparent glitch reveals that despite improvements, these systems still struggle with elementary tasks that humans find trivial. The error was noticed by TechCrunch and has since circulated as an example of AI’s ongoing challenges with factual accuracy.
This incident underscores a recurring concern in AI development: as language models become more sophisticated and are deployed more prominently in user-facing applications, they continue to make embarrassing mistakes on fundamental tasks. The issue is particularly notable because it occurs in a product Google positions as an enhancement to search accuracy. The company would presumably prefer such errors remain hidden rather than displayed prominently to millions of search users. While the error is superficially amusing, it raises legitimate questions about the reliability of AI-generated content in critical applications and the current limitations of LLM technology when handling concrete factual information.