A new study from Columbia Journalism Review showed that AI search engines and chatbots, such as OpenAI's ChatGPT Search, Perplexity, Deepseek Search, Microsoft Copilot, Grok and Google's Gemini, are just wrong, way too often.