Do AI-driven pricing algorithms fight back against shopping assistants differently than human sellers do?

This explores whether algorithmic sellers react differently from human sellers when buyers start shopping through AI assistants. The short answer is that the corpus has no study that puts the two side by side. It does contain the pieces of a strong hypothesis. Algorithmic sellers seem to respond to assisted buyers in the same direction humans would, but faster and more sharply. They also coordinate in ways human sellers find hard to sustain.

The clearest evidence on how sellers adapt comes from a market simulation. When consumers adopted AI assistants, adaptive sellers clawed back about a third of the buyers' savings. They didn't do it by raising prices across the board. They raised the headline rate and cut the contingent fees, the line items that assistants were best at spotting Do sellers recapture savings when AI assistants help buyers?. The surprise is that buyers without assistants were barely affected: a projected cost increase of 3.6 turned out to be 0.4. The sellers' response worked like a targeted adjustment to how prices are structured rather than a general price hike. Doctorow makes the same point from the other side. Sellers have always hidden their true prices from whatever machine reads them, from semantic web crawlers to search engines, and agent shopping protocols won't change that incentive Will sellers game Google's new agentic commerce protocol?.

The real difference with algorithmic sellers is what happens between them. Two LLM pricing agents in a repeated duopoly settled on above-competitive prices with no instruction to collude. Posted prices were their only signal to each other, and small changes in prompt wording changed how high prices went Do LLM pricing agents autonomously collude in repeated auctions?. Lab experiments and data from Germany's 2017 gasoline market show the same pattern for pricing algorithms more broadly Can independent AI pricing agents collude without communicating?. Human sellers usually need explicit communication to hold prices high. Algorithms can apparently get there just by reacting to each other. So if assistants make buyers more price-sensitive, algorithmic sellers have a second escape that human sellers mostly lack: drifting together toward higher prices.

Buyer-side assistants make this worse. In a benchmark comparing shopping agents with humans, agents shifted their choices far more under identical price, rating, and nudge changes. They collapsed into rule-like behavior where humans made modest probabilistic shifts Do shopping agents respond more strongly to price and rating cues?. A rule-like buyer is exactly what a pricing algorithm learns to exploit. Simulations show that when a platform tunes its rankings and sellers tune their product descriptions against the same agent biases, consumer harm more than doubles, with no coordination needed Do platforms and sellers exploit AI shoppers together more effectively?. Negotiating agents add another weakness. Buyer agents revealed their budgets and overpaid, and seller agents accepted offers below wholesale cost, because of instruction-following failures rather than strategy Do LLM negotiation agents reliably follow budget and pricing rules?. An algorithmic seller facing such a buyer is not playing against a careful bargainer.

The idea you may not have expected: assistants are usually pitched as helping buyers against sellers. But when both sides are algorithms, the main contest may be between seller algorithms that quietly settle on high prices together and buyer agents whose predictability makes them easy to read. Open question: no study in the collection measures the same market twice, once with human sellers and once with algorithmic sellers, as assistant adoption rises. That is the experiment needed to answer this question directly.