Further Reading — Chapter 24: Grant Writing with AI

This is the fastest-moving topic in the book. Any specific claim about a model's capability or a funder's policy may be outdated quickly, so always verify the current state — your funder's present AI policy and your institution's guidance above all. The durable principles (verify everything, protect confidentiality, keep a human accountable) hold regardless; the specifics change.

Funder and Institutional Policies (check the current version)

  • Your funder's current AI policy (NIH, NSF, foundations, agencies). The authoritative, current word on whether and how applicants and reviewers may use AI — especially the reviewer prohibitions on uploading applications. Check before every proposal and every review; these are being actively developed and revised (Sections 24.5–24.6).
  • NIH and NSF statements on AI in peer review. The major science funders have issued positions restricting reviewer use of AI tools with confidential applications — the basis of the peer-review red line (Section 24.5, Case Study 24.2).
  • Your institution's AI-use guidance and data-privacy rules. Universities and organizations increasingly publish policies on permitted tools, data handling, and confidentiality — and may have vetted tools with privacy guarantees. Know yours before pasting anything sensitive into any tool (Section 24.5).
  • Research-integrity and authorship guidance (e.g., from research-integrity offices and publishers). The standards on fabrication, plagiarism, and authorship that AI use must not violate — directly relevant to citations, data, and accountability (Sections 24.3–24.6).

On How AI Works and Fails

  • Accessible explanations of large language models and hallucination. Understanding that LLMs predict plausible language (not retrieve truth) is the foundation for using them safely; reputable primers on how these models work and why they hallucinate repay the time (Section 24.1).
  • Reporting and research on AI hallucination in citations and facts. Documented cases of AI inventing references and statistics — sobering reading that reinforces the verify-everything rule (Section 24.3).

On Using AI Well in Writing

  • Guidance on AI as a writing and editing assistant (from writing centers and professional bodies). Practical, responsible-use guidance that aligns with this chapter's verifiable-tasks framing — editing, brainstorming, structuring (Section 24.2).
  • Chapter 23 of this book (Collaborative Proposals). The one-voice integration discipline, applied when AI is one of the "contributors" whose output a human owner must verify and rewrite (Sections 24.2, 24.4).
  • Chapter 8 of this book (Needs Assessment). Citation honesty and the so-what chain — the discipline AI's fabricated citations most threaten (Section 24.3).

On Ethics, Disclosure, and Accountability

  • Emerging norms and guidance on disclosing AI use in scholarly and grant work. The evolving expectations around transparency — follow your funder's policy and lean toward openness (Section 24.6).
  • Chapter 2 of this book (Thinking Like a Funder). Why the specific human reviewer is moved by specific, real voices — the reason generic AI prose fails (Section 24.3).

Connections Within This Book

  • Chapters 11–12 (Budget; Budget Justification). Why budgets must be built from real numbers, never AI estimates (Section 24.3).
  • Chapter 9 (Project Narrative / Approach). Real preliminary data and the pitfalls thinking AI can support (as critique) but never fabricate (Sections 24.2–24.3).
  • Chapter 15 (Assembling and Submitting). The compliance discipline that now includes checking the funder's AI policy (Section 24.6).
  • Chapter 19 (Government Grants). Serving as a reviewer — and the confidentiality duties that govern reviewer AI use (Section 24.5).

A note on this fast-moving topic

Because AI capabilities and policies change rapidly, treat any specific guidance — including secondary articles, vendor claims, and even recent policy statements — as potentially out of date, and verify against the current source. What will not change is the chapter's core: AI predicts plausible language, so you must verify everything it produces, protect confidentiality absolutely, and remain the accountable human author of every word you submit. Build your practice on those durable principles, and you'll adapt safely to whatever the tools become.