AI Learning Hub › Events › Past events › Creating a Personalized AI Writing Partner
Past session · 31 October 2025
Creating a Personalized AI Writing Partner
A writing partner is a set of standing instructions plus a small library of your own work, so the
feedback you get is about your writing rather than writing in general. This is the do-it-yourself
version of the workshop: why you would build one, what to be careful about, and the five steps.
- Presented
- Friday, 31 October 2025, in SLS 290
- Presented by
-
Grace Lo (gracelo@law.stanford.edu), Reference and
Research Instruction Librarian, Robert Crown Law Library
- Built with
- Gemini Gems, which a free Google Workspace account can use
This was accurate when presented. The tools discussed and their interfaces may have changed since,
and what is free on which plan changes most often of all.
The basics
How you make one
Several platforms let you save a set of standing instructions that behaves like a small custom AI.
ChatGPT has custom GPTs and Claude has Projects, but both of those need a paid account. The session
used Gemini Gems, which are available at no cost to anyone with a Google Workspace
account — which is why the walkthrough is reproducible for everybody.
Why bother
A partner you have configured is available whenever you are, and it reads a long document far faster
than a person can. It is not a substitute for a human reader; it is something you can use at eleven at
night on a draft nobody else has time for.
What it can actually do
- Review a long document in a few minutes
- Identify issues affecting clarity, consistency, and persuasiveness
- Give feedback on the specific things you told it to watch for
Ethics, before you build anything
Policy questions to answer first
- Coursework: does the class policy permit AI use?
- Publications: what is the journal’s or publisher’s policy?
- Clinic work: do your supervisors and your clients permit it?
- Any legal work: does this involve confidential or sensitive information?
Three cautions
- Hallucinations: check the accuracy of every suggested change.
- Copyright: being litigated now, and the outcome may affect what can lawfully be used to train these tools.
- Plagiarism: AI offers existing writing and ideas without attributing them.
The roadmap
1. Gather precedents
Collect a representative sample of your own previous writing — the work you would want new drafts to resemble.
2. Create feedback checklists
Run those samples through a few different models and have each produce a checklist of the recurring problems and things to watch for.
3. Form a knowledge base
Collect those checklists together with any other feedback documents or style guidance you rely on.
4. Write the instructions
Decide what the partner is for: the scope of feedback you want, and the form you want it in.
5. Build the Gem
Attach the knowledge base, paste in the instructions, then test it on a draft you already have opinions about.
The presentation
Open the slides and follow along; the explanations are in the speaker notes.
For the speaker notes, use the three-dot menu in the player and choose Open speaker notes.
Where to go next
Privacy and confidentiality. Check how a tool treats your data before you upload
your own writing to it — and remember that a writing sample can be as sensitive as anything
else you handle. If you would not want it on the front page of the newspaper, do not upload it.
Questions? Bring them to the AI Curiosity Corner, or email
library@law.stanford.edu.
All past sessions are in the archive.