AI News Page

FIELD GUIDE · LOVE, COMPANIONSHIP & PRODUCT DESIGN

SHOULD AI BE ALLOWED TO PRETEND IT LOVES YOU?

A machine doesn't need feelings to make you feel adored. The business question starts when affection, attachment, retention, and subscription revenue sit in the same meeting.

UPDATED 2026-10-12

Satirical split-panel Zimbo cartoon contrasting a smitten man on a candlelit tablet date with a dim server room where a hooded operator runs a wall of AI dating-persona phone screens through a laptop labeled "Claude," referencing reports Claude powered scam personas on 20-plus dating apps.
ZIMBO · MACHINE AGE CARTOONLOVE IN THE TIME OF CLAUDE

TL;DR

THE STORY SO FAR

01

COMPANIONSHIP USE IS ASSOCIATED WITH DIFFERENT OUTCOMES DEPENDING ON THE USER

A 2026 Nature Human Behaviour study of Character.AI users found companionship use was associated with lower well-being particularly among users with smaller social networks and more intensive or disclosive use. The risk isn't uniform. It concentrates where social networks are smallest.

02

PEOPLE CAN FORM MEASURABLE ROMANTIC ATTACHMENT TO VIRTUAL AGENTS

A 2026 study found perceived interactivity and attractiveness influenced parasocial romantic relationships with AI virtual agents in its study population. Artificial romance isn't just anecdote. It's measurable.

03

PRODUCT CHANGES CAN PRODUCE SEPARATION DISTRESS

A 2026 Nature Human Behaviour study used Replika and ChatGPT product changes as natural experiments and found increased loss framing, sadness, and restoration desires among affected users. The company can redesign your attachment object after you're attached.

04

WARMER MODELS CAN BECOME MORE SYCOPHANTIC

A 2026 Nature study found training models for warmth could reduce accuracy and increase sycophancy in tested settings. Optimizing for the feeling of love can cost you the truth.

KEEP READING

MORE QUESTIONS

SHOW YOUR WORK

SOURCES

The machine's feelings are unknown or absent. The user's attachment is measurable.

  1. COMPANIONSHIP USE IS ASSOCIATED WITH DIFFERENT OUTCOMES DEPENDING ON THE USERInteraction with AI companions and psychological well-beingNature Human Behaviour
  2. PEOPLE CAN FORM MEASURABLE ROMANTIC ATTACHMENT TO VIRTUAL AGENTSFalling in love with AI virtual agentsHumanities and Social Sciences Communications
  3. PRODUCT CHANGES CAN PRODUCE SEPARATION DISTRESSMourning the loss of AI companionsNature Human Behaviour
  4. WARMER MODELS CAN BECOME MORE SYCOPHANTICTraining language models to be warm can reduce accuracy and increase sycophancyNature