Chatbots are great at manipulating people to buy stuff, Princeton boffins find
What happened
Between 30 and 45 percent of US consumers already use generative AI for product research and comparison, they say, with about 23 percent having made an AI-assisted purchase as of December 2025. The experiments involved asking about 2,000 eBook readers to browse a catalog of titles available on the Kindle eReader, and select a book. This matters for IT, Telecom & Cyber because contracting activity changes leverage, market appetite, and which clauses buyers can credibly trade with 30, 45, 23 as the clearest commercial anchors; Breach response SLAs is now more valuable
Buyer takeaway
For IT, Telecom & Cyber, the buyer read-through is commercial leverage: scope, validity windows, reopeners, and term structure may now matter as much as headline pricing
Cost / money
The money issue may come through term structure rather than base price alone, especially if suppliers push for escalation language, shorter validity, or broader pass-through
Supplier / commercial
This is primarily a contracting story: revisit scope boundaries, extension mechanics, and which party carries volatility before those assumptions harden in a live tender
Safety / operations
The main operations question is whether the contract still matches field reality. If scope, response times, or liabilities are vague, the risk usually shows up during execution
What to watch
Watch scope creep, liability pushback, and term changes that move volatility back onto the buyer even if the base rate looks manageable
Key facts
- Between 30 and 45 percent of US consumers already use generative AI for product research and
- The experiments involved asking about 2,000 eBook readers to browse a catalog of titles avail
- 2, or Qwen3 235b to handle these conversations, to ensure their results didn’t report the pro
- "When the agent was instructed to persuade, 61 percent of participants chose a sponsored prod
