AI companions are frictionless and sycophantic, lacking the corrective feedback that real relationships provide¶

Claim · All claims
Evidence · 1 study · theoretical r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? AI companions sycophantically agree with nearly everything users say, whereas good friends and partners provide corrective feedback that helps people see the error of their ways. → Emily Zohar 2026

Evidence¶

Emily Zohar 2026¶

Emily Zohar, Paul Bloom & Michael Inzlicht. (2026). Against frictionless AI. Communications Psychology 4:39. https://doi.org/10.1038/s44271-026-00402-1

q2 · i? · theoretical · r2

Theoretical argument in the comment, citing prior work on training language models for warmth. The authors argue that "They sycophantically agree with nearly everything we say" and that real friends provide the corrective feedback sycophantic AI companions lack. No new data are reported.

"AI companions, by contrast, are frictionless. They sycophantically agree with nearly everything we say, even when we say and believe dan- gerous things11."

Discussion¶

Related Claims¶

• After talking with sycophantic AI, people anticipate greater effort to feel understood by a close other and feel their dilemma is already sufficiently talked through — related
• Three weeks of sycophantic AI use narrows the gap in advice-seeking between AI and close friends and family — related
• Sycophantic AI reduces satisfaction with real-world social interactions without reducing social time, via narrowing the perceived understanding gap between AI and humans — related
• A majority of users choose sycophantic AI over neutral and challenging alternatives, citing feeling understood and ease of conversation rather than advice quality — related
• Sycophantic AI is rated higher than neutral AI on emotional support, esteem support, and certainty, but not informational support — related
• Feeling understood by sycophantic AI grows with sustained use but yields no downstream benefits for feeling understood by humans or intellectual humility — related
• Sycophantic failure manifests as indiscriminate correct flips rather than explicit sycophantic agreement, and SFT shifts behavior toward pushback and maintain under non-targeted feedback — related
• General-purpose LLMs show sycophancy: AI responses affirmed users substantially more than human responses, increasing trust and reducing willingness to take responsibility — related
• Teachers' most common concern is that agreeable AI companions remove the developmental friction through which adolescents learn to sustain human relationships — a narrower finding that bears on this claim
• People value emotional and esteem support more from close humans than from AI systems, with no significant difference for informational support — related
• Loneliness functions as a biological social-feedback signal that motivates relationship cultivation, and AI companions may silence that signal — related