> For the complete documentation index, see [llms.txt](https://docs.aicrisk.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.aicrisk.com/user-documentation/getting-started/user-chat-page/agent-reasoning-and-citations.md).

# Agent Reasoning and Citations

**Agent Reasoning**\
ChatGPT 5.4 includes a **reasoning dropdown** that lets you view a high-level summary of how the model arrived at its answer. This section highlights key steps, assumptions, and factors considered—without exposing full internal chain-of-thought—so you can better understand the logic behind the response.

**Citations**\
When applicable, responses include **citations** that point to supporting sources (e.g., web pages, documents, or data). These appear inline and help you:

* Verify accuracy
* Explore sources in more detail
* Distinguish grounded information from general knowledge

Together, the reasoning summary and citations improve transparency and trust in model outputs while keeping explanations clear and usable.

<figure><img src="https://2203016705-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FAFZNv6i2feb00nkSIqAC%2Fuploads%2FNCbaLO21DdgiShJxXoMP%2Fimage.png?alt=media&amp;token=8091ca0a-de0b-4ebc-b2ba-ea061c7a2203" alt=""><figcaption></figcaption></figure>