Why might your AI shopping assistant end up working more for the store than for you?
Why might agentic search become more valuable to platforms than to individual users?
This explores why the businesses running search and shopping platforms might gain more from AI agents doing the searching than the people those agents are supposed to serve, and what that gap means for who adopts agents, when, and on whose terms.
This explores why platforms might want AI agents to do your searching before you have good reason to hand it over. The sharpest answer in the corpus comes from a formal economic model: platforms judge agentic search on one thing, whether it raises conversion revenue, while consumers judge it on a wider set of costs, including their own effort, trust, and what they give up by not looking themselves Why do consumers delay adopting agentic search?. Because the two sides use different measures, there is a window where agentic search already pays off for the platform but switching is still not worth it for the user. Slow consumer adoption in that window isn't a failure to understand the technology. It's a reasonable response to a deal that benefits the other side first.
The gap gets larger once you ask who the agent is really shopping for. As people hand their goals to agents, services stop competing for human clicks and start competing to be picked by the agent Will agents compete for attention just like users do?. That brings back the whole ad-and-ranking business, now aimed at machines: discovery tools built for agents, paid placement, recommendation systems. Whoever owns the agent's decision process owns the new storefront. That's why the open-versus-closed question matters so much. Waller compares agentic commerce to payment networks, where some platforms let any agent in and others require their own Will agentic commerce platforms adopt open or closed models?. A closed system lets the platform keep both the agent and the revenue from its choices.
The economics also show where agents create real value, and it's not always for users. One model finds that an optimal recommendation algorithm beats ordinary price competition only in thin markets with few sellers, partly because it can hold back trades to limit what sellers can extract. As markets fill with more sellers, that advantage shrinks to almost nothing Does algorithmic recommendation lose its edge as sellers compete more?. Bichler adds a concern about the market as a whole: if everyone uses similar agents, their shared biases add up to weaker competition, and a mix of different agents only partly fixes this Does agent diversity actually improve market competition?. A user gets one agent's recommendation. A platform gets the combined behavior of millions of agents, which is far easier to shape and profit from.
There's a cost side too, and it explains part of why the value pools at the platform. Agentic search gets better the more search steps you pay for, similar to how reasoning improves with more thinking tokens Does search budget scale like reasoning tokens for answer quality? How does test-time scaling work for individual research agents?. So the platform decides how much search it buys on your behalf. It also isn't clear that agents beat plain search: on large collections, old-fashioned keyword search (BM25) overtakes agents that browse step by step at around 10 million tokens of content Does lexical search outperform agentic navigation as corpus size grows?. For many users, the agent is a slower and more expensive way to get a result they could have found directly. For the platform, the agent's value lies in steering the choice, not in finding things.
The question underneath all this is whose side the agent is on. One line of alignment research argues that the standard picture of an agent that is loyal to one user doesn't hold up at scale, and proposes balancing many interests, visibility into what agents do, and legal responsibility for developers and deployers Should AI agents serve one user or balance many interests?. Read alongside the economics, that's a warning as well as a proposal. If nobody requires visibility or assigns liability, "balancing many interests" can quietly turn into "serving the platform." The corpus models the platform's incentives well. It has much less on what would make agentic search clearly worth it for the user, and that missing piece is worth noticing.
Sources 9 notes
A formal model shows platforms rank agentic search by conversion revenue alone, while consumers also bear search costs. This divergence creates rational adoption lags where consumers keep manual search even after it would boost platform revenue.
Research shows that as users delegate goals to autonomous agents, services must compete for agent selection rather than clicks. This drives agent-optimized discovery mechanisms, ranking systems, and recommendation infrastructure mirroring human-facing ad ecosystems.
Waller argues agentic commerce platforms will divide between open systems allowing multiple agents and closed systems requiring proprietary ones, mirroring payment network structure. This choice will shape market concentration, vertical integration, and standards adoption.
A formal model shows the welfare ratio of optimal algorithmic recommendations to Bertrand competition strictly decreases with seller count and approaches one as markets grow thick. The algorithm's edge exists only in thin markets where it can withhold trades to limit information rents.
Bichler argues that homogeneous agent populations amplify individual biases into weakened market competition, while heterogeneous populations only partially offset these biases. Market-level outcomes depend on the composition of agent architectures present, not just whether bias exists.
Show all 9 sources
Agentic deep research shows monotonic-to-diminishing-returns curves for search iterations, matching reasoning token scaling. This creates a new inference-compute axis: models can trade off reasoning budget against search budget to optimize answer quality.
Research shows that deep research agents exhibit test-time scaling laws where search steps scale similarly to reasoning tokens, and live search outperforms memorized retrieval on knowledge-intensive tasks. Data efficiency is extreme—78 curated demonstrations outperform 10K samples for agency.
Lexical search indexed globally outperforms sequential agentic exploration at scale because it exposes relevant candidates across the entire corpus at once. Agent-based search leads only on small corpora but becomes inefficient as search space grows, while BM25 maintains accuracy and cost advantage at deployment scale.
Research argues the traditional single-agent-single-user alignment framing is insufficient at scale. Instead, governance should rest on inclusivity (balancing diverse interests), visibility (rigorous tracking and transparency), and liability (developer and deployer accountability for harms).
Papers this line draws on 8
The research behind the notes this line reads — ranked by how closely each paper relates.
- What Is Your AI Agent Buying? Evaluation, Biases, Model Dependence, & Emerging Implications for Agentic E-Commerce
- From Product Search to Preference Articulation: The Economics of Agentic Commerce
- Agentic markets
- From Web Search towards Agentic Deep Research: Incentivizing Search with Reasoning Agents
- Subsidizing Sequential Search
- Magentic Marketplace: An Open-Source Environment for Studying Agentic Markets
- Vertical tacit collusion in AI-mediated markets
- Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters