Chatbots powered by artificial intelligence provide ready-made answers to everything in an instant, but the information is not always correct, according to a test conducted by the British consumer protection organization «Which?». Six AI-powered chatbots were tested with 40 questions in finance, law, health, nutrition, consumer rights, and travel. Experts then evaluated the quality of the answers. The test results showed that most chatbots perform well, but none can be trusted blindly. Perplexity bot achieved the best performance with 72% correct answers, followed by Google's AI Mode at 70%, then Google Gemini at 68%, Microsoft Copilot at 67%, and ChatGPT at 65%. Meta AI bot ranked last with 54%. According to the test results, the AI sometimes ignored regional differences or recommended risky or potentially illegal behavior, while other answers relied on questionable or outdated sources, overlooked free alternatives to paid services, or downplayed consumer rights. Here are three tips for using AI answers more safely:
— Carefully review every AI answer, especially when it comes to important financial, legal, or health topics. — Always ask the AI about its sources and consult another expert. — Phrase questions for AI chatbots as precisely as possible and provide as much basic information as possible; this significantly improves the answers.
