Emotion AI in marketing, after the ban at work
Emotion recognition and affective computing appear on 325 recent papers, almost all in AI journals; only 19 carry a consumer, marketing or ethics keyword. The EU now bans emotion recognition at work and in schools, but not in shops or advertising. Does reading faces and voices tell marketers anything reliable, and should it be allowed?
What prompted this#
Since October 2025, 325 papers carry an emotion-recognition, affective-computing, facial-expression or eye-tracking keyword (301 in AI journals): "emotion recognition" on 76, "multimodal emotion recognition" on 60, "affective computing" on 53, "facial expression recognition" on 29. Only 19 also carry a consumer, marketing, advertising, retail, privacy or ethics keyword (neuromarketing 4, AI ethics 3).
Questions#
- Do emotion scores from faces or voices predict purchases or ad recall better than simply asking people, out of the lab and across cultures?
- How accurate are commercial emotion AI tools for groups underrepresented in their training data?
- Since the EU banned emotion recognition in workplaces and education, have vendors moved to retail and advertising, and do consumers know they are being read?
Data that could help#
Field experiments in advertising testing with and without emotion measures, vendor accuracy audits by demographic group, and vendor product and market data before and after the EU rules.
Next step#
Compare how well facial emotion scores and short self-reports predict choices among the same participants in an online ad-testing experiment.