Sponsored Question Answering
Tommy Mordo, Moshe Tennenholtz, Oren Kurland
TL;DR
The paper investigates how to migrate from sponsored search to sponsored QA by proposing a formal auction-based platform that fuses an organic answer with ads to generate sponsored answers. It formalizes a GSP-like mechanism where advertisers bid on sponsored answers, and the winner is selected to maximize a platform value that combines user utility with advertiser bid; bidding truthfully emerges as a dominant strategy, yielding VCG-like welfare properties. Because end-to-end fusion with advanced LLMs is complex, the authors analyze a tractable unigram-language-model framework to study the auction's behavior, showing that the highest bidder need not win and that user utility is not always maximized. The work provides a foundational theoretical framework for sponsored QA, highlighting welfare considerations and guiding design choices, while acknowledging ethical and practical challenges in blending organic and promotional content. Overall, it offers a first formal treatment of sponsored QA auctions and their economic implications for both users and advertisers.
Abstract
The potential move from search to question answering (QA) ignited the question of how should the move from sponsored search to sponsored QA look like. We present the first formal analysis of a sponsored QA platform. The platform fuses an organic answer to a question with an ad to produce a so called {\em sponsored answer}. Advertisers then bid on their sponsored answers. Inspired by Generalized Second Price Auctions (GSPs), the QA platform selects the winning advertiser, sets the payment she pays, and shows the user the sponsored answer. We prove an array of results. For example, advertisers are incentivized to be truthful in their bids; i.e., set them to their true value of the sponsored answer. The resultant setting is stable with properties of VCG auctions.