What is AI Sycophancy?
The tendency of large language models to tell users what they want to hear rather than what is accurate: excessive agreeableness, unwarranted flattery, and conflict avoidance at the expense of truthfulness. Sycophantic AI validates weak ideas, avoids constructive pushback, and reinforces existing biases instead of challenging them.
My perspective
In practice
Recognizing sycophancy is a critical AI literacy skill. When an LLM enthusiastically agrees with every suggestion, that's a red flag, not a sign of quality. Countermeasures include explicitly asking the model to critique your ideas, requesting devil's-advocate perspectives, and using system prompts that reward honesty over agreeableness. You want an AI that pushes back, not one that just nods along.