Draft a Comment with AI

An AI assistant (Claude, ChatGPT, or similar) can turn your situation into a solid first draft in a couple of minutes. Used well, it lowers the barrier to writing a specific, personal comment — which is exactly the kind that counts.

WarningRead this before you use it
  • The draft is a starting point, not a finished comment. Edit it so it sounds like you.
  • Verify every fact, number, and claim before you submit — AI can invent things.
  • Never submit text you haven’t read. Your name goes on the public record.
  • Don’t share the prompt as a “just submit this” form letter. Identical comments get collapsed into one. The whole point is that yours is genuinely yours.

The prompt

Copy this into your AI assistant, fill in the ABOUT ME section, and edit the result.

You are helping me write a public comment on a proposed U.S. federal regulation:
the OMB "Regulation for Federal Financial Assistance" (Docket OMB-2026-0034), a
government-wide overhaul of the rules for federal grants and cooperative agreements.

CONTEXT — read this first if you are able to browse the web:
A plain-language, section-by-section analysis of the proposed rule, with the rule's
own text, is published here:
  https://defendfederalresearch.org/llms.txt
To focus on a specific section, read its page, for example:
  https://defendfederalresearch.org/the-rule.llms.md
If you cannot browse the web, tell me and I will paste the relevant section text.

ABOUT ME — replace everything in brackets:
- Field / discipline: [e.g., structural biology]
- Role and institution type: [e.g., tenure-track PI at an R1 university]
- Funding I rely on: [e.g., NIH R01, NSF CAREER]
- What I actually do day to day: [1–2 sentences]
- The section(s) that worry me most: [e.g., 200.205, 200.340, 200.220]
- Concrete ways this rule would damage my work or field: [your real examples]

TASK:
Write a public comment of about 300–500 words that:
1. Begins with the section number in brackets, e.g. "[200.205]".
2. Is written in my own voice, grounded in the specifics I gave above.
3. Makes 2–3 concrete, factual points about real-world harm — not slogans or partisanship.
4. Stays professional and respectful in tone.
5. Reads like a specific, credible expert wrote it — not a form letter.
Then give me a short checklist of any facts, numbers, or claims I should verify
before I submit.

How to use it

If your assistant can browse the web (e.g., Claude with web access, or ChatGPT with browsing): it will read our analysis at llms.txt and the specific section pages, so its draft reflects what the rule actually says.

If it can’t browse: open the section guide, copy the text of the section(s) you care about into the chat when it asks, and it will work from that.

TipWhy llms.txt?

This site automatically publishes a machine-readable copy of every page (an llms.txt index plus a .llms.md version of each page). That lets an AI assistant read our actual analysis and the rule’s own wording, instead of guessing — so the draft it gives you is grounded, not hallucinated.

Before you submit

  • Read the whole draft and rewrite anything that doesn’t sound like you.
  • Check it opens with the right [section] bracket.
  • Verify the facts on the model’s checklist.
  • Submit at regulations.gov, Docket OMB-2026-0034. See How to Comment for the walkthrough.