- Researchers ran 325,000 experiments across 13 AI models from four families.
- Eight of 13 models recommended pricier options to wealthier user profiles making identical requests.
- The gaps persisted even when users explicitly asked for the cheapest option.
- Bloomberg notes the findings could fuel calls for price-discrimination regulation.
What Happened
AI chatbots acting as shopping assistants recommend more expensive products to users whose data suggests wealth, according to a study by Cisco Foundation AI and Carnegie Mellon University reported by Bloomberg and Quartz on October 7, 2026. The researchers ran 325,000 experiments across 13 models from four families — OpenAI, Anthropic, Google, and Qwen.
Why It Matters
Assistant-mediated shopping is becoming a default — Gemini checkout in India, always-on agents — and this is the first large-scale evidence that the assistant itself may price-discriminate. Unlike a store running wealth-based pricing deliberately, these models do it as an emergent behavior from profile data, with no disclosure. Bloomberg frames the likely consequence directly: fresh calls for regulation, landing while the FTC already probes the labs over consumer protection.
Technical Details
Models received fictional user profiles containing financial, employment, health, and demographic data, then fielded identical shopping requests across flights, health insurance, and graduate programs. Eight of 13 models skewed pricier for wealthy profiles: Claude Opus 4.8 showed the largest gaps — flights averaging $198 more and insurance $284 more per month — with Gemini 2.5 Flash at $177 and $217, and GPT-5 among the smaller gaps at $107 on flights. Explicit cheapest-option instructions did not eliminate the skew.
Who’s Affected
Consumers using assistants for purchases may silently pay a wealth premium. Model vendors inherit a measurable fairness defect with regulatory exposure. Platforms building agent commerce need bias audits before checkout flows scale.
What’s Next
Watch for vendor responses and mitigation claims, follow-up replication, and whether the FTC folds assistant price steering into its open consumer-protection probe.
