# How we built ChatGPT's novel ad stack for user trust and relevance

DevFeed: [How we built ChatGPT's novel ad stack for user trust and relevance](<https://devfeed.tech/articles/how-we-built-chatgpt-s-novel-ad-stack-for-user-trust-and-relevance-92879.md>)

Original publisher: [Read original article](<https://openai.com/index/building-chatgpt-ad-stack>)

Published: 2026-10-21T07:00:00Z

Content type: article

Language: en

Sources: [OpenAI News](<https://devfeed.tech/sources/openai-news.md>)

Topics: [ChatGPT](<https://devfeed.tech/topics/chatgpt.md>), [AI Chat](<https://devfeed.tech/topics/ai-chat.md>), [AI interaction best practices](<https://devfeed.tech/topics/ai-interaction-best-practices.md>), [UI/UX Design](<https://devfeed.tech/topics/ui-ux-design.md>), [trust](<https://devfeed.tech/topics/trust.md>)

Tags: [ads](<https://devfeed.tech/tags/ads.md>), [advertising](<https://devfeed.tech/tags/advertising.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-loop](<https://devfeed.tech/tags/agentic-loop.md>), [ann](<https://devfeed.tech/tags/ann.md>), [annualized-revenue](<https://devfeed.tech/tags/annualized-revenue.md>), [approximate-nearest-neighbor](<https://devfeed.tech/tags/approximate-nearest-neighbor.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [chatgpt](<https://devfeed.tech/tags/chatgpt.md>), [engineering](<https://devfeed.tech/tags/engineering.md>), [learn](<https://devfeed.tech/tags/learn.md>), [llm](<https://devfeed.tech/tags/llm.md>), [novel](<https://devfeed.tech/tags/novel.md>), [openai](<https://devfeed.tech/tags/openai.md>), [relevance](<https://devfeed.tech/tags/relevance.md>), [scale](<https://devfeed.tech/tags/scale.md>), [separate](<https://devfeed.tech/tags/separate.md>), [stack](<https://devfeed.tech/tags/stack.md>), [trust](<https://devfeed.tech/tags/trust.md>), [understand](<https://devfeed.tech/tags/understand.md>), [user](<https://devfeed.tech/tags/user.md>)

## AI overview

ChatGPT's ad stack uses a custom post-trained LLM orchestrator to interpret evolving conversational intent and call separate tools for ad retrieval, eligibility checks, relevance and engagement scoring, and auction selection. It maintains state across turns, can consider nonsensitive past-conversation context when users opt into ad personalization, and may return no ad when none meets its constraints. Ads run in a separate service from answer generation, so the article says ads cannot influence responses or delay them if the ad service is slow or unavailable. Even a selected ad may not appear if it feels inappropriate in context or arrives after the response is ready; displayed ads are labeled sponsored, with a neutral preamble kept separate from the advertiser's creative. The system uses detailed outcome logs and user and advertiser feedback to guide improvements.

## Source excerpt

Learn how OpenAI built ChatGPT's LLM-powered ad stack to understand conversational intent, keep ads separate from answers, and deliver relevant ads at scale.