The invention of Retrieval-augmented generation (RAG) has dramatically changed how complicated an AI's response can be, and how much more visibility a user can have into its logic.
Now, instead of simply summarizing a topic or a primary source, AI can collect information from multiple sources and aggregate it into a single response. Citations help users trace the information contained in a response back to its original material.
This can be within a single source, the way that Adobe PDF and other document summarizers point to different sections, or across multiple references, akin to Wikipedia-like footnotes.
![](https://cdn.prod.website-files.com/65db6dd21591364dfcb8ae36/6681cdbba817374a76ee648d_rag.png)
The details differ a bit from platform to platform. Copy.ai shows the full url of the reference and no preview. Perplexity highlights the top references (users can also edit the primary references to regenerate the response). Because adobe only references a single source in its summary, it's actions all point to paragraphs within the document itself.
Citations help users go deeper into a topic, whether tracing through internal documentation, conducting discovery during research, or trying to verify certain information. They offer a mechanism to cover the mundane work of aggregation so users can focus on editing and understanding.
Citations are commonly found in summaries (e.g. timestamps on summarized video transcripts) and in synthesized responses (like Perplexity's multi-reference responses). AI Chatbots and Agents may also provide references to show what information they are referencing as they work through their tasks.
As AI aggregation is increasingly relied upon by consumers for discovery and research, citations provide a tie back to the original content, benefitting content creators with traffic to their site and benefitting consumers with footprints of information.