If AI agents start doing our shopping and searching for us, who do businesses have to impress now — us, or the agent?

This explores whether, as AI agents take over more of our shopping, searching and comparing, the thing businesses compete for shifts from human attention to something else, and whether the companies that run agents end up holding that new scarce resource. The short version from the corpus is that competition doesn't go away. It moves somewhere else, and where it settles depends on choices about platforms that are still being made.

The case that the shift is real is fairly strong. Once people hand a goal to an agent, services have to win the agent's selection instead of a click. The corpus predicts this will produce a whole new layer of agent-facing ranking, discovery and recommendation systems that mirror today's ad ecosystem Will agents compete for attention just like users do?. Some early evidence is already in. Jarc argues that when an agent ranks hotels or products by the criteria a person actually stated, paid placement stops mattering, because the agent keeps the 'demand signal' (what the buyer wants) rather than handing it to an aggregator. Jarc links Tripadvisor's 23% drop in Hotels revenue to this pattern Are AI agents making attention-based ads irrelevant for aggregators?. So the old model of buying a human's glance does weaken when no human is glancing.

The less obvious point is that the scarce resource may not be 'agent attention' at all. It may be control over which agents get to act and what they can see. Waller compares agentic commerce to payment networks and expects a split between open systems, where any agent can participate, and closed ones that require a proprietary agent Will agentic commerce platforms adopt open or closed models?. Rothschild and colleagues warn that Apple, Google, Meta and Microsoft are well placed to rebuild app-store-style gatekeeping at the layer where agents connect to each other Will agent economies develop as open webs or closed gardens?. The UK's competition regulator (the CMA) adds that if people can't take their data with them, their agent can't act well on their behalf and switching becomes impractical. In other words, lock-in replaces attention as the source of market power Should data portability be infrastructure for agent-driven markets?.

That raises a loyalty question that the attention economy never had to answer this directly: whom does your agent actually serve? Economics and agency law already name the problems of hidden information, discretion and divided loyalty. They also note that the usual fixes, such as incentives, monitoring and enforcement, break down when an agent makes fast decisions nobody can easily interpret Can traditional governance tools actually control AI agents?. If sellers start paying to sway agent rankings, the ad economy comes back inside the agent, where it's harder to see. That's one reason researchers argue alignment has to look beyond pleasing a single user to visibility and liability across everyone affected Should AI agents serve one user or balance many interests?. It's also why some propose internet-style infrastructure that records which agent did what and can detect harm and fix it What infrastructure must exist beyond training agents?.

So the corpus's answer is a qualified yes. Human attention is losing its place as the thing everyone fights over. But what replaces it is less 'agent attention' than two other things: a position at the platform layer that decides which agents get in, and the trust that an agent is really working for you. This collection doesn't yet have hard data on whether the agent layer will end up open or concentrated. That remains the open question, and the most consequential one.