Further Reading — Chapter 29: Writing with AI

Annotated, Tier 1 and Tier 2 sources only. A deliberate warning for this chapter: AI is a fast-moving field, and the specific models, benchmarks, and "studies" shift month to month. I have not cited capability statistics, named studies, or product comparisons here — they would be stale or fabricated. What's durable is the writing craft beneath the tool and the integrity frameworks around it, so that's what this list points to. Treat any specific claim about "what AI can do" — including in this book — as provisional and check the current state yourself.

Tier 1 — The writing craft the tool can't replace (the durable foundation)

William Strunk Jr. & E. B. White, The Elements of Style. The cure for the AI default register. "Omit needless words" is the exact remedy for the fluent, empty, "revolutionary-seamless-cutting-edge" prose a model produces when you ask it to supply substance it doesn't have (§29.4). When you edit an AI draft, you're applying Strunk and White to it — re-read the conciseness sections with a fluent AI paragraph in front of you.

William Zinsser, On Writing Well. Zinsser's "clutter is the disease of American writing" describes the generic AI voice precisely, and his insistence that good writing comes from clear thinking is Chapter 1's thesis and this chapter's core: the model can produce the words, but the thinking — the part Zinsser cares about — is yours and can't be delegated.

Joseph M. Williams, Style: Lessons in Clarity and Grace. The deep treatment of why sentences read clearly — the principles you use to judge and revise an AI's output. The whole skill of using AI as a revision tool (§29.4) depends on having a clear standard for what "better" means; Williams gives you that standard.

Tier 1 — Citation, integrity, and verification (the §29.6 foundation)

Your institution's or organization's official AI-use policy. The single most important "source" for this chapter, and the one most readers skip (§29.6). Academic-integrity offices, journals, and many employers have published explicit guidance on disclosure and acceptable AI use. It is specific to your context, it is authoritative for you, and it answers the disclosure question this chapter could only answer in general. Read it before you use AI on anything that goes out under your name.

The citation-style authorities (APA, the Chicago Manual of Style, IEEE, MLA). Most major style guides have published guidance on how to cite or disclose AI-generated content. As with Chapter 11, the durable point isn't the exact format (which is being revised as norms settle) but the principle: AI use, where it touches the substance, is something readers may need disclosed — check the current edition of the style your venue requires.

Tier 2 — Frameworks and communities (real, widely held; cited honestly)

The verification discipline (Chapter 11, this book). The three-rule framework — disclose, verify, don't present as yours what you can't evaluate — is this book's own §8 practice and Chapter 11's treatment, applied to AI. It's the most directly useful thing to re-read alongside this chapter, because §29.6 and §29.7 are its logical extension.

The "plausible, not true" framing of language models. The understanding that LLMs predict likely text rather than retrieve verified facts is widely held among practitioners and researchers and is the foundation of §29.1. The specific technical literature changes fast and is beyond this book's scope; the durable idea — optimize for plausible, indifferent to truth — is the part you need, and it's the part least likely to be outdated.

Documented real-world AI failures (e.g., sanctioned legal filings citing fabricated cases). Several widely-reported instances exist of professionals submitting AI-hallucinated citations and facing consequences (the §29.6 and Case Study 2 example). These are real and instructive; search for current cases rather than relying on a fixed list, since new ones appear regularly — each is a live demonstration of the governing rule violated.

A note on the citation tiers (this book's §8 practice): I have deliberately omitted the kind of source this chapter is about — specific AI capability studies, benchmark numbers, named recent papers, and product comparisons — because in a field moving this fast they would be stale or, worse, the very hallucinations the chapter warns against. The durable sources are the writing-craft classics (Tier 1) and the integrity frameworks (Tier 1/2). For the current state of the technology, go to primary, dated sources and verify — and apply the governing rule to anything you read, including this book.

Back to: Chapter 29 · Key Takeaways · Exercises