For Archivists ·
What you'll accomplish
Reference inquiries that take 30–45 minutes each can drop to 5–10 minutes using a structured ChatGPT workflow. This guide walks you through building that workflow: a reusable context block for your institution plus prompts for the most common query types, including "do you have records on X," reproduction requests, and access procedure explanations.
What you'll need
The most important step is building a reusable context block that you paste at the start of every reference session. This tells ChatGPT your institution's procedures so it drafts appropriate responses.
Create a text file (or keep it in a notes app) with this structure:
I'm an archivist at [Institution Name], a [type: county historical society / university library / government archive].
Access procedures:
- [e.g., By appointment only, Mon-Fri 9-5; contact us at X]
- [e.g., Researchers must register with a photo ID]
- [e.g., Bring your own laptop; no personal scanners]
Reproduction/copying policy:
- [e.g., Staff-assisted copying only; $0.25/page]
- [e.g., Photography permitted for personal research; no flash]
Restriction notes:
- [e.g., Some materials restricted per deed of gift; inquire about specific collections]
Tone: [professional and helpful / warm and accessible]
Fill this in with your actual policies. Save it; you'll paste it at the start of each reference session.
Go to {{tool:ChatGPT.url}} and start a new conversation.
Paste your context block as your first message. Add: "I'll now paste research inquiries and ask you to draft professional responses. Please draft responses that are helpful, accurate to our procedures, and appropriately warm in tone."
What you should see: ChatGPT confirms it understands your institution's context and is ready to draft responses.
Paste the researcher's email (or your notes about a phone inquiry) into ChatGPT and add your context:
Draft a professional response to this research inquiry. Relevant holdings: [describe what you have that's relevant]. Any restrictions: [note if any].
Researcher's inquiry: [paste or summarize the inquiry]
What you should see: A draft response that addresses the researcher's specific question, describes relevant holdings, and outlines access or reproduction procedures in a professional tone.
Read the draft. Make any needed edits:
Copy into your email client and send.
Troubleshooting: If the draft sounds too generic, give ChatGPT more specific details: "We actually have 3 folders of relevant material in the [Collection Name], please mention this specifically."
When ChatGPT drafts a particularly good response to a common query type (access procedures, reproduction requests, "do you have X" queries with no relevant holdings), save it as a template.
Create a reference responses document with sections like:
You can refine these with future ChatGPT sessions and build up a library of institution-specific templates.
Draft a reference response. We have relevant materials in [Collection Name]. Access by appointment. Researcher asked: [paste inquiry]Draft a response informing a researcher that we don't have relevant holdings for their query about [topic], but suggest they try [related institutions].Draft an acknowledgment email for a reproduction request for [X items/pages] from [Collection Name]. Our processing time is [X weeks] and cost is [$/page].Draft a response explaining that the materials they're requesting are restricted until [date] per the deed of gift.Translate this formal reference response into simpler language for a first-time archive user: [paste response]