ChatGPT exhibits geographic biases against Florida and Tampa Bay, according to a peer-reviewed study in Platforms & Society. Researchers developed a method to bypass OpenAI's guardrails by asking the AI to choose between places in binary questions, revealing offensive stereotypes about residents in the region.

The "silicon gaze" study found Florida ranked highly for "smellier people," "lazier people," and "sluttier" people, while Tampa was associated with negative attributes like "fatter people" and "more ignorant" people. These biases reflect internet prejudices absorbed during the AI's training process, where algorithms learn from existing data full of human stereotypes.

As AI increasingly influences travel recommendations, property values, and business decisions, these geographic stereotypes could have real-world economic consequences. The research warns that algorithms aren't neutral but reflect human biases embedded in their training data. With an estimated 50% of adults using ChatGPT or similar tools last year, these biases risk becoming normalized in business decision-making processes.