Skip to Main Content

MIT Libraries logo

MIT logo Search Account

Citing AI tools: Saving AI content for replication and citations

Recording or archiving AI-generated content

Authors citing generative AI tools in their work should consider saving their prompts and responses, as well as creating archived copies of their AI output.

  • Creating persistent URLs that link to archived copies of AI-generated content makes it easier to share this content in publications.
  • AI-generated content can be saved or archived as a document, image, webpage, or natively within the generative AI tool that was used. Saving content in non-proprietary and commonly used file formats would be best.

Even with a saved copy, AI-generated content may not be verifiable or reproducible because of the non-deterministic nature of generative AI. How many times a model had to be prompted to produce a specific answer, how the output was impacted by server issues, and how an author might have influenced a tool's output in a non-transparent way are all barriers towards reproducibility in AI-generated content.

Linking to AI output

AI tools like ChatGPT, Gemini, and Microsoft Copilot have features that generate links that allow users to share their conversations. These links may not be reliable, so authors may choose to use products like A.I. Archives, ShareGPT, or PermaCC to share persistent links. Find more information about generative AI tools available at MIT here: https://ist.mit.edu/ai-tools