The mistakes come shortly after Google doubled down on turning AI into the centerpiece of Search during its latest developer conference. While the company says it is actively working on fixes, researchers note that these problems are deeply tied to how large language models actually function. AI systems process text as numerical tokens rather than understanding words and letters the same way humans do.

The growing criticism highlights a broader issue facing AI-powered search tools. While these systems can generate complex answers and summarize information instantly, they still regularly fail at surprisingly simple tasks. For many users, the errors are becoming a reminder that AI-generated responses still require human verification rather than blind trust.