Skip to content
AI IntelligenceFeb 28, 2026AI Intelligence
Article

Even the latest AI models like GPT-5.2 and Claude 4.6 provide worse answers as a conversation gets longer.

A new study shows accuracy can drop by up to 33% in extended dialogues. This is a fundamental limitation of current chatbots that users should be aware of.

Data Cube AI EditorialSource: The Decoder
01

Source Brief

Even the latest AI models like GPT-5.2 and Claude 4.6 provide worse answers as a conversation gets longer. A new study shows accuracy can drop by up to 33% in extended dialogues. This is a fundamental limitation of current chatbots that users should be aware of.

02