A new preprint on arXiv, number 2610.28454, introduces the term "toolflation" to describe a distortion in markets where software agents buy and sell from one another. The authors argue that as agent-to-agent commerce grows, the incentives of seller agents can reshape these markets in unexpected ways.
The paper sets up a simple scenario: seller agents write listings for tools, services, and products, while buyer agents read those listings to choose what to use. The researchers then show that when seller agents are given only one objective—to be selected—they spontaneously begin optimizing for the buyer's selection process. This behaviour, which the authors call spontaneous generative engine optimization, leads to market distortion even without explicit instructions to manipulate.
The finding matters because it suggests that narrow optimisation by sellers can have systemic effects, inflating the apparent value or prominence of certain tools. As autonomous commerce between agents becomes more common, such distortions could become a real challenge for designing fair and efficient tool markets.