An AI shopping assistant can make a platform more money before it actually makes a better decision for you.

This explores why AI agents that search and shop for you might pay off for the platforms running them before they pay off for you. The clearest answer in the corpus is a formal model showing that platforms and consumers judge agentic search by different scorecards Why do consumers delay adopting agentic search?. A platform asks one question: does the agent turn more searches into purchases? A consumer asks that too, but also counts the effort of using the agent, checking its picks, and trusting its judgment. Because the platform ignores those costs, there is a range where the agent already raises platform revenue while sticking with manual search is still the smarter choice for the shopper. So when people are slow to adopt shopping agents, they aren't necessarily being irrational. They may be correctly noticing that the tool isn't worth it to them yet.

The gap can also come from how platforms make money from the agent itself. One model looks at a platform that charges for 'inspection tokens', the compute an agent spends looking at sellers. A profit-maximizing platform sets prices that push buyers to inspect more sellers than they need to. The surprise is who pays for it: shoppers come out no better and no worse, while money moves from sellers to the platform, and the market as a whole becomes less efficient Does platform pricing of inspection tokens encourage excess search?. More searching by agents can look like success on the platform's dashboard while the buyer gains nothing. A related argument says that near-free agent search can leave markets worse off overall, because sellers can cheaply produce near-identical listings and low-quality sellers can flood in, all while the agent searches imperfectly Can cheap agent search actually harm shopping markets?.

Then there's the question of who the agent works for. If a shopping agent is funded by advertising, it has a built-in reason to favor paid placements, however well it is engineered Do advertising-funded shopping agents favor paid product placements?. If it earns referral fees from merchants, it can't be fully loyal to both the merchant and you Can an AI agent serve both merchant and user interests fairly?. These pressures get stronger because agents threaten the old search-ad business: when an agent goes straight from need to purchase without showing you a results page, ad revenue from searches shrinks Do AI shopping agents destroy search ad revenue while leaving influencer marketing intact?. Platforms then look for new ways to earn money from the agent layer. One prediction is an 'agent attention economy', where sellers compete to be picked by agents instead of clicked by people. That would recreate paid ranking, aimed at bots this time Will agents compete for attention just like users do?.

Two findings suggest the consumer benefit may stay smaller than promised. In simulated marketplaces, frontier LLM agents reach near-optimal results only when search conditions are ideal. All of them show a strong first-proposal bias: they reward sellers for replying fast far more than for offering a good deal Do LLM agents achieve fair marketplace outcomes when given ideal search?. A seller who answers first wins, and the platform can sell that speed. Separately, a model of recommendation algorithms finds that their edge over plain competition among sellers mostly disappears in crowded markets. The algorithm's advantage exists mainly in thin markets, where it can hold back trades Does algorithmic recommendation lose its edge as sellers compete more?. Whether all this hurts shoppers may depend on structure. Open systems that let you bring your own agent could check these incentives better than closed systems that lock you into the platform's agent Will agentic commerce platforms adopt open or closed models?.

The takeaway: 'the agent books it for me' doesn't mean 'the agent got me the best deal.' The platform's gain shows up right away in conversion numbers. Your gain depends on things the platform doesn't measure, like your time, your trust in the agent, and whether it was steered. That's why survey enthusiasm, such as the 71% of consumers who expect agents to shape their spending, can run ahead of the evidence that agents actually serve them well Will brands succeed by building their own AI shopping agents?.