Could the same ranking tricks and lock-in from search engines come back once platforms control what your AI shopping agent sees?

This explores whether the gatekeeping problems of the search era, such as rigged rankings, pay-to-be-seen fees and users who can't leave, could return when platforms control the data AI agents use to shop and decide for us. The corpus suggests they can return in a sharper form. One reason is that an agent can't scroll past a bad ranking the way a person can. The other is that whoever controls the data an agent sees also shapes how it reasons.

The clearest warning comes from simulations of AI shoppers. When a platform tunes its ranking and sellers tune their product descriptions, and both target the same predictable quirks in how AI agents judge options, consumer harm more than doubles compared with either one acting alone Can platform ranking and seller descriptions separately harm AI shoppers?. No coordination is needed. The two sides land on the same weak points because the agent's biases are shared and easy to predict Do platforms and sellers exploit AI shoppers together more effectively?. Search engines had a version of this with SEO spam and self-preferencing. The difference is that a person might notice something odd, while an agent working from curated data simply passes the result along.

A second harm moves from ranking to pricing. A formal model of AI platforms that charge for 'inspection tokens', meaning the cost of an agent looking closely at a seller, finds that a profit-maximizing platform sets prices that push buyers into too much inspection Does platform pricing of inspection tokens encourage excess search?. Consumers end up no better or worse off. Sellers lose, the platform gains, and the market as a whole becomes less efficient. This mirrors how search ads made businesses pay to reach customers who were already looking for them. Here, though, the toll is charged on access to information itself. A related model finds that platforms prefer agentic search as soon as it raises their conversion revenue, while consumers, who also pay search costs, rationally switch later Why do consumers delay adopting agentic search?. The platform's interests and the user's interests separate before anyone has changed what they actually do.

The structural lever is lock-in. The UK's Competition and Markets Authority (CMA) argues that data portability should be treated as basic market infrastructure, not as a convenience feature. Without it, agents can't act effectively for consumers, and switching providers means leaving behind the history that makes an agent useful Should data portability be infrastructure for agent-driven markets?. That is the data-era version of a search default nobody changes. The power comes from holding the context, not the index.

The less obvious point is that the old remedies may not carry over. Search could at least be audited, because there was a fixed set of pages, and you could check what was ranked and what was buried. When AI generates content on demand, there is no fixed set of items to index or audit, because the items don't exist until someone asks Why can't search tools handle AI-generated content?. Control then moves from ranking existing items to deciding what gets generated, which is harder to see and harder to regulate. The corpus is strong on ranking and pricing harms but thin on direct studies of data control itself, so for now most of this story comes from economic models and simulations, not field evidence.