@techreport{TR-IC-PFG-26-10,
   number = {IC-PFG-26-10},
   author = {Filipe Lacerda Benevides and Julio Cesar dos Reis},
   title  =  {{Sponsored  Discovery  Mechanism  for  the  Web  of Open
                   Agents}},
   month = {July},
   year = {2026},
   institution = {Institute of Computing, University of Campinas},
   note = {In English, 31 pages.
    \par\selectlanguage{english}\textbf{Abstract}
       As  autonomous  {LLM}  agents  become  primary consumers of web
       services,  they  discover  and  delegate to one another in open
       ecosystems  --  a  \emph{Web  of  Open  Agents} -- yet existing
       discovery infrastructures offer no principled way to govern how
       agents   \emph{compete   for  visibility},  risking  pay-to-win 
       dynamics   in   which  advertising  spend,  not  merit,  drives 
       selection. We propose \emph{Sponsored Discovery}, whose core is
       {MACS}  (Match-gated  Agent  Composite  Score): semantic domain
       match  acts  as  a \emph{multiplicative gate} that structurally
       bounds  out-of-domain  agents  regardless of bid or reputation,
       coupled  with  a  multi-dimensional  behavioral reputation over
       accuracy,  latency,  and  cost.  Using  real  {LLM} agents that
       answer  {MMLU-Pro} questions across three adversarial scenarios
       (honest       baseline,       domain-falsifying      fraudster,     
       {SLA}/cost-inflated   agent),   {MACS}   achieves  the  highest 
       delegation  success across all three. Our results from a paired
       ablation   show  {MACS}'s  advantage  over  a  purely  additive 
       composite:   \emph{decisive}   under  domain  misreporting  and 
       \emph{marginal} in benign settings, where behavioral reputation
       alone  suffices,  while bid-only and quality-adjusted baselines
       are  exploited.  This  study  further  analyzes  its  incentive 
       properties  and  limits  (calibrated misrepresentation, {Sybil}
       whitewashing).
  }
}