If one seller's AI rewrite can steal market share, what happens when every seller does it at once?

This explores what happens when many sellers, not just one, use AI agents to rewrite their product listings to win over AI shopping agents. The collection doesn't have a study that tests this directly. It does have the single-seller result that sets up the question, plus nearby findings that suggest how the multi-seller version might go.

Start with the single-seller case, which is surprisingly strong. In the ACES experiments, one seller's agent read competitor listings and rewrote only its own product description. It didn't touch price, rating or position, yet it gained 3.66 to 14.89 percentage points of market share with five of the six buyer models tested Can AI seller agents shift buyer preferences through description edits?. So AI shoppers respond to wording in a way that can be measured and exploited. But market share is zero-sum. Those points came from rivals who weren't optimizing. Once every seller runs the same playbook, a reasonable guess (an inference, not a finding in the corpus) is that the gains largely cancel out. You'd get an arms race in persuasive wording, where descriptions are tuned for the buyer model's quirks rather than for informing anyone.

The closest evidence about agents pursuing separate goals in a shared space is not encouraging. When each user had their own personal agent in a shared-resource environment, the group did consistently worse than when a single coordinator served everyone. In settings with no way for the agents to communicate, the teams collapsed entirely Do teams of personal agents outperform a single coordinator?. A marketplace of seller agents is that same setup: many self-interested agents and no coordinator. It's a plausible model for how a description arms race could degrade the whole market, not just shuffle the winners.

There's also a twist that cuts against the idea of a sophisticated strategic contest. When LLM buyer and seller agents negotiated, they often broke their own rules. Buyers revealed their budgets and overpaid, and sellers accepted offers below wholesale. Those failures traced back to agents not following instructions, not to clever strategy Do LLM negotiation agents reliably follow budget and pricing rules?. So a multi-seller market might be less like chess and more like noise: lots of description rewriting, with outcomes driven by each buyer model's blind spots rather than by which product is better.

The part you may not have thought to ask about is the human at the end of the chain. If transactional shopping agents help shape what consumers want through repeated back-and-forth, rather than just carrying out preferences people already had Do AI shopping agents reshape what consumers actually want?, then a contest among seller agents is about more than market share. It indirectly shapes what buyers end up preferring, through text written mainly for other machines to read. If you want to go deeper, the open question the corpus leaves is whether buyer agents can be made robust to that pressure, or whether the market needs something like a coordinator.