AI in Search Engines: The Good, the Bad, and the Ugly

This paper examines the integration of artificial intelligence (AI) models in search engines, exploring their accuracy, reliability, and ethical implications. Testing popular AI-enhanced search engines and AI models, including Microsoft’s Copilot, Google’s AI Overview, DuckDuckGo’s AI Chat, and Perplexity, this paper compares their performance in answering factual and reasoning-based questions. Results revealed that while AI models generally provide accurate responses to straightforward queries, they often struggle with complex or reasoning-based questions. Solutions proposed to address these challenges include providing enhanced user control, improved ethical standards for AI data training, separation of traditional search engines and AI models, and promoting the use of alternative search engines and tools outside of the commercially used products. These solutions aim to improve user trust, transparency, and information reliability in AI-enhanced or powered search products. The findings are a call to action to encourage a balanced approach to integrating AI in search engines, prioritizing user choice, ethical considerations, and accuracy to create a more dependable search experience.