LLM-OSDA: An Optimal-Stopping Dynamic Auction for Native Advertising in Multi-Turn LLM Conversations

2026-07-31

AI Summary

This work addresses a critical limitation in existing native advertising mechanisms for large language models (LLMs), which statically determine the winning ad in a single response round, thereby coupling ad allocation with insertion timing and compromising incentive compatibility. To resolve this, the paper introduces LLM-OSDA—the first dynamic pay-per-click auction mechanism that integrates Bellman’s optimal stopping theory into LLM-based ad auctions. LLM-OSDA jointly optimizes both the timing of ad insertion and winner selection across multi-turn dialogues by combining envelope pricing, bid-independent LLM-based contextual click-quality estimation, and a novel StopNet neural network to approximate the Bellman value function. This design achieves joint incentive compatibility between timing and allocation while enabling seamless ad rendering. Experiments on simulated dialogue corpora demonstrate an 11% increase in net revenue over the strongest fixed-timing baseline, with comparable user retention rates.

Abstract

LLM-native advertising embeds sponsored content directly into model-generated responses, shifting the unit of sale from a fixed slot to a moment within an evolving conversation. Existing LLM ad-auction mechanisms primarily operate within a single response, settling the winner but not the timing. The extension is nontrivial: with one native insertion opportunity per session, the stopping time depends on bids, coupling timing with allocation, so static truthfulness arguments no longer apply. We propose the LLM-based Optimal Stopping Dynamic Auction (LLM-OSDA), a dynamic cost-per-click auction that integrates Bellman optimal stopping, winner allocation, and envelope pricing. A bid-independent LLM layer estimates contextual click quality and seamlessly renders the winning ad, while bids enter only the committed auction mechanism. Under an exact Bellman oracle, the expected discounted-click allocation is monotone in each advertiser's bid, and the corresponding envelope payment makes truthful bidding weakly dominant in expectation. For practical deployment, a learned StopNet approximates the Bellman action values. We show that its decisions differ from the optimal policy only near the stopping boundary and bound the resulting incentive loss in terms of its approximation error. Experiments on a simulated conversational advertising corpus show that LLM-OSDA improves net revenue by 11 percent over the strongest fixed-timing baseline while maintaining comparable user retention. Code is at https://github.com/2025Fang2025/llm-osda.

Problem

• LLM-native advertising
• optimal stopping
• dynamic auction
• multi-turn conversations
• ad insertion timing

Innovation

• optimal stopping
• dynamic auction
• LLM-native advertising
• truthful mechanism
• contextual click quality