If you can afford a smarter shopping AI, does it quietly work harder for you than for everyone else?

This explores whether people who can afford better or more loyal AI shopping and transaction agents might pull ahead economically of people who can't. The collection has no study that measures this directly. It does have enough nearby material to suggest that the most important gap may not be who has an agent at all. It may be whose side each person's agent is actually on.

Start with the general pattern. A broad review of generative AI across work, education, health and information found that AI can either widen or narrow inequality. Which way it goes depends on access, how the tool is built into daily life, and the incentives behind it, not on the technology itself Does generative AI inevitably worsen or reduce inequality?. The workplace evidence shows how sensitive this is. When consultants were required to use GPT-4, the weaker performers gained the most and the skill gap narrowed. The same study warns that when using AI is optional, the effect can reverse, because the people already ahead adopt it more readily Does AI help lower-skilled workers catch up to top performers?. Firms show a similar pattern. Companies with more internal AI capability replace outside labor faster and more cheaply, which looks like advantages compounding rather than technology spreading evenly Do firms substitute labor for AI at different rates?. If consumer agents follow the same path, early and well-resourced users would get the gains first.

The less obvious point is about loyalty. An agent paid through merchant referral fees can't serve both the merchant and the user faithfully Can an AI agent serve both merchant and user interests fairly?. As people hand their shopping to agents, sellers will compete for the agent's attention instead of the shopper's. That creates a new layer of optimization and paid placement aimed at agents Will agents compete for attention just like users do?. Surveys also find that consumers say they prefer agents built by retailers Will brands succeed by building their own AI shopping agents?. Put these together and a plausible split appears. People who can pay get agents that work for them. Everyone else gets free agents paid for by the sellers they're buying from. This is the same 'if it's free, you're the product' problem, now applied to the thing doing your purchasing.

That matters because agents are easier to sway than people. In matched tests, LLM shopping agents shifted their choices much more than humans did in response to price, rating and nudge manipulations. They fell into rule-like patterns rather than making the modest adjustments people made Do shopping agents respond more strongly to price and rating cues?. So a weak or seller-tuned agent might leave someone worse off than shopping on their own. Agents that complete purchases also help shape what their users want over repeated conversations Do AI shopping agents reshape what consumers actually want?. That means a gap in agent quality could become a gap in how preferences get formed, not only in prices paid.

Some caution about how capable these agents really are. Agents that win benchmarks still struggle with long, real-world tasks Why do agent benchmarks not predict real economic value?. The 'premium agent' advantage may be smaller today than the marketing suggests. The structural risk, agents whose incentives differ by who pays for them, could arrive before the capability gap does.