Updated weekly

AI Pulse

The most important developments in humanized AI — companions, emotional intelligence, neuromarketing, and human-AI interaction research. Curated, not aggregated.

ai-companions

Character.AI Surpasses 100 Million Monthly Active Users

Character.AI has crossed 100 million monthly active users, cementing it as the dominant AI companion platform globally. The milestone signals that parasocial relationships with AI personas are moving from novelty to daily habit — with teenagers averaging over an hour per day on the platform, raising both opportunity and ethical questions about emotional dependency.

emotional-ai

MIT Study: AI Can Reliably Detect Emotional Dysregulation Before Humans Do

A new MIT Media Lab study found that a multimodal AI model — analyzing voice prosody, word choice, and typing cadence — detected emotional dysregulation in subjects 4–7 minutes before trained therapists identified the same signals. The research opens the door to real-time mental health support tools that could intervene during emotional crises before they escalate.

neuromarketing

The Neuromarketing AI Wave: Agencies Start Replacing Eye-Tracking Labs with Models

Several mid-size advertising agencies have begun retiring their physical neuromarketing labs — fMRI rigs, eye-tracking hardware, galvanic skin sensors — in favour of AI models that predict attention, arousal, and emotional response from creative assets alone. Vendors like Realeyes, Neurons Inc., and emerging startups are driving the shift, with reported cost reductions of 80% versus traditional biometric testing.

ai-companions

Replika Enters Clinical Trial as Adjunct Therapy Tool for Loneliness

Luka, the company behind Replika, has partnered with the University of Southern California to run a randomised controlled trial testing Replika as an adjunct tool for treating chronic loneliness in adults over 55. If results are positive, it would be the first AI companion app to earn a clinical evidence base — a move that could reshape how insurers and regulators treat emotional AI products.

human-ai-interaction

Stanford HAI: Perceived Warmth — Not Accuracy — Drives Long-Term AI Trust

A Stanford Human-Centered AI Institute study of 3,200 participants found that users who rated an AI assistant as 'warm' or 'caring' continued using it six months later at three times the rate of users who rated it as merely 'accurate.' The finding upends the assumption that performance is the primary loyalty driver in human-AI interaction, and suggests emotional design deserves equal engineering investment as capability.