AI-Assisted Ghostwriting: How It Works for Business Books

A ghostwriter’s proposal arrives, and on page three there is a line you did not expect: “I use AI tools in my process.” Or its opposite: “Every word is written by a human.” Neither sentence tells you much. Both affect your fee, what happens to your interview recordings and what you have to tell Amazon when the book goes live.

The trade only began writing shared rules down in April 2026, so most authors sign without asking the questions that matter. This piece is for both people on that call: the coach or consultant hiring help, and the ghostwriter or book coach doing the work. If you are still deciding whether to hire anyone at all, start with our comparison of an AI book writer and a ghostwriter.

What ghostwriters actually reported about AI

The most useful evidence comes from the trade itself. In summer 2025, Gotham Ghostwriters and the writer and analyst Josh Bernoff surveyed 1,481 working writers for a study called “AI and the Writing Profession”. Publishers Weekly reported that 61% used AI tools, that users claimed an average productivity gain of 31%, and that only 7% had published AI-generated text.

Bernoff then broke out the 115 respondents who are primarily book ghostwriters. Of these, 68% used AI at least sometimes. Among the weekly users, the top tasks were search (77%), other research (66%), brainstorming (63%) and titles and headings (62%). A quarter used the transcription tool Otter.ai. Only 5% generated text with AI and published it. Bernoff warns against overreading 115 responses (margin of error 9%).

The conference floor was blunter. In the American Society of Journalists and Authors’ write-up of the 2025 Gathering of the Ghosts (Michelle Rafter, 14 November 2025), panellists advised starting with low-value, time-consuming jobs such as transcription, research and title brainstorming. According to the write-up, every ghostwriter who used AI said they did not use it for the writing itself. Some still transcribe interviews themselves to absorb the author’s voice. Michael Long, a Georgetown lecturer, put the risk of AI prose in one line: “Everything you leave out will be filled in with blandness.”

Treat this as a stated norm among people on stage, not a census. Still, it fits the survey, and it fits why you hire a ghostwriter: the sentences are the product.

Three working models you are likely to be offered

“AI-assisted” covers very different arrangements. Most proposals describe one of three.

Model	What AI does	What the human does	Likely KDP status of the text
1. AI around the work	Transcribes interviews, searches, checks spelling	Interviews, structures and writes every sentence	Human-written; AI-assisted at most
2. AI organises the material	Summarises transcripts, groups stories by theme, suggests outline options, lists gaps	Chooses the structure, writes every sentence	Human-written; AI-assisted
3. AI first draft, human rewrite	Drafts chapters from transcripts and notes	Rewrites heavily, fact-checks, restores the author’s voice	AI-generated, even after heavy edits

The working group convened by Gotham Ghostwriters published AI guidelines for ghostwriting in April 2026. According to Publishers Weekly’s report, its “Recommended Disclosures” are grouped the same way: administrative tasks; “research, analysis, and collaborations”; and generative uses such as “generating initial drafts of text content that will be revised later”. That gives you and your writer a shared vocabulary.

Our view: Model 2 is where AI earns its keep on a business book, because a coach’s material is scattered across years of decks, recordings and half-finished articles, and sorting it is slow work. Model 3 is legitimate only when it is disclosed and priced as what it is. A writer who drafts with AI and calls it Model 1 is the arrangement to avoid.

A hypothetical example: an executive coach gets three proposals. Writer A transcribes with Otter and writes everything. Writer B builds a story bank and gap list from the transcripts with an AI tool, then writes. Writer C rewrites AI drafts of each chapter. All three are “AI-assisted”. Only C’s book needs a KDP disclosure, and only B and C put her client stories into an AI system beyond transcription. She needs those two facts before comparing fees.

What changes in price and timeline, and what does not

Start with the anchor. Reedsy’s marketplace data (updated 2 June 2026) puts business and management ghostwriting at $30,000 to $48,000, about $0.73 a word. One ghostwriter quoted there uses AI to transcribe interviews but checks every transcript, because “AI can’t manage every accent.” For the full fee picture, see our guide to business book ghostwriter cost.

We could not find published data showing that AI-assisted ghostwriting is priced lower. The closest signal is income. In Bernoff’s breakdown, ghostwriters who use AI had a median income only 14% higher than those who do not ($116,300 against $102,500), compared with a 45% gap across all writing professionals. A reasonable reading is that AI trims hours from transcription and research, not from the interviewing, judgement and rewriting that set a ghostwriter’s fee.

So for Models 1 and 2, do not expect a discount. For Model 3, ask what the saved hours become: a lower fee, a shorter schedule or more revision rounds.

Your own hours barely move under any model. You still sit for the interviews and review every chapter, and AI cannot make you reply to a draft any sooner, which is where most schedules slip. To shorten the interviews, prepare with our guide to extracting your knowledge before writing a book.

Your interview recordings are the most sensitive part of the project

A business book interview is full of things you would never put on LinkedIn: client names, revenue figures, the engagement that went wrong. In Models 2 and 3 that material goes into an AI system; in Model 1 it at least goes to a transcription service.

The Authors Guild’s AI best practices for authors (May 2026) is direct about this: “By default, material you enter into almost all public chatbots can be used to train those models,” and uploading drafts, outlines, research or “interview transcripts may expose confidential material.” It adds that it is not fully clear how opt-out settings will be honoured over time.

Account type matters. When Anthropic updated its consumer terms in August 2025, chats on its Free, Pro and Max plans became usable for training when the user’s setting is on, with retention extended to five years in that case, while products under its commercial terms, such as Claude for Work and the API, were excluded. Other providers draw their own lines, and those lines change. A writer on a personal chat account and one on a business account with training off offer you very different protection.

The Gotham guidelines list five AI risks, according to Publishers Weekly: possible lack of copyright protection, confidential material in training data, plagiarism, inaccurate transcription and factual errors. For protecting your own frameworks, see protecting your original ideas when writing with AI.

Putting AI use into the contract

This is not legal advice. Have a lawyer read any agreement that matters commercially.

The Authors Guild recommends that authors who hire a ghostwriter, co-author, editor or book coach sign written contracts that “prohibit the addition of more than de minimis AI-generated text without your written approval” and “require that they disclose their use of AI”. It also notes that publishing contracts include a warranty that the manuscript is original, and “AI-generated material is not considered ‘original authorship’”. In a traditional deal, your writer’s undisclosed AI text could become your breach.

We would not ask for a blanket “no AI” clause. Spellcheckers, search engines and transcription apps all use AI now, so a total ban invites quiet non-compliance. List the permitted uses instead:

• Permitted uses by category: admin, research and analysis, generative. Generative use needs your written sign-off.
• Named tools and settings: which services process your material, on what account type, with training off where the option exists.
• Retention: what is deleted at the end of the project, and by when.
• Ownership: you own the manuscript and the work product, including transcripts, story banks and notes produced with AI.
• Accuracy: the writer flags every claim, figure and quote they could not verify from your material.
• Disclosure at handover: a short written note of how AI was used, chapter by chapter if Model 3 was involved.

Disclosure runs both ways. If you drafted notes with ChatGPT before the project started, tell your writer; they carry the same risk of invented facts. Our guide on stopping AI inventing facts in your book explains what to check.

What KDP disclosure means for a ghostwritten book

Amazon’s KDP content guidelines, as of November 2026, require you to inform Amazon of AI-generated content: text, images or translations. Text an AI tool created counts as AI-generated “even if you applied substantial edits afterwards”. AI-assisted content, where a person created the text and used AI to edit, refine, error-check or brainstorm, does not need disclosing.

The test is who created the text, not who you paid. Under Models 1 and 2 a person writes it, so it is AI-assisted at most. Under Model 3 the book contains AI-generated text however good the rewrite. KDP also makes you, the account holder, responsible for verifying that AI-generated and AI-assisted content meets its guidelines. For a self-published book, that is you, not your ghostwriter.

Where it is unclear, we would lean towards disclosing; getting it wrong costs more reputation than ticking the box. Our guide to publishing a business book on Amazon KDP covers the rest of the upload process.

Questions to ask your ghostwriter about AI

Ask these on the first call, before any fee discussion. A good ghostwriter will not mind.

We would treat vague answers to the third and fourth questions as a red flag. On voice, the more of your own writing a ghostwriter has, the less any tool can flatten it; see training AI on your writing style for which samples to collect.

If you are the ghostwriter or book coach

Clients are starting to ask these questions, so answer them first. Publish a plain description of your AI use, map it to the three categories above and put the agreed uses in every contract. That is roughly what the 2026 guidelines ask of writers anyway, and it turns a nervous conversation into a selling point.

For production work, guided book software is one option alongside general chat tools. WritebookAI is one example. It builds a Writing DNA voice profile from the client’s own samples, which the client tests and approves. It analyses documents, notes and talk transcripts to show what they cover and miss, runs a verification interview on the gaps and builds a chapter plan. It drafts chapters from that material; when a client result or source is missing, the chapter leaves it out and asks rather than inventing it. It does not replace your interviews, your judgement or a human copyedit, and because it drafts text, the result is Model 3 and needs a KDP disclosure. Check its privacy policy against your client agreements as you would with any tool. Details are on the WritebookAI homepage; our guide to AI tools for ghostwriters and book coaches covers the process from intake to handover.

Before your first call with a ghostwriter

Take three things into the conversation: a one-page note of what is confidential (client names, numbers, anything under NDA), your preferred model from the table above, and the checklist.

Then compare proposals on what each writer will actually do, not on whether the word “AI” appears. If none fits your budget or the way you work, our guide to writing a business book without a ghostwriter sets out the other routes.