HOT RADAR
llmHEAT 15.5°

AI CURATED HOT STORY · 10/21/2026

OpenAI Details ChatGPT Ad Stack: Stateful Retrieval and Auctions via LLM Tool Calls

1 reports archived1 independent sourcesupdated 10/21/2026, 15:00:00
Synthesis & Latest Updates
1 Sources Cross-Validated

OpenAI revealed a bespoke ad stack for ChatGPT powered by a post-trained frontier LLM that uses tool calls for retrieval, ranking, relevance evaluation, and auctions. The system introduces 'stateful ad sessions' to track evolving user intent across conversation turns, aiming to balance user experience with advertiser value.

LATEST/OpenAI revealed a bespoke ad stack for ChatGPT powered by a post-trained frontier LLM that uses tool calls for retrieval, ranking, relevance evaluation, and auctions. The system introduces 'stateful ad sessions' to track evolving user intent across conversation turns, aiming to balance user experience with advertiser value.

HEAT TRENDHourly heat curve

3 fully observed hours
Now
15.5
Peak
16.210-22
24h change
–

The line compares only sources observed throughout; gap hours are interpolated to keep the trend continuous.

TIMELINECoverage timeline

Total 1 reports · Latest first
  1. OpenAI 官方博客T1·68 pts
    • Ad decisions are executed by a post-trained LLM via tool calls covering retrieval, ranking, relevance evaluation, and auctions
    • Innovative implementation of 'stateful ad sessions' allows the system to remember and interpret evolving user intent across multiple turns and conversations
    • The architecture features self-improving capabilities through continuous feedback loops to optimize personalization relevance
OpenAI Details ChatGPT Ad Stack: Stateful Retrieval and Auctions via LLM Tool Calls | Today for AI