The same ghost, a different haunt: practical steps for human-centred ghostwriting with AI

The ghost and the machinery

We begin with a simple observation from the original piece: ghostwriting has always been both social and technical. Scribes, collaborators and professional ghostwriters have long produced text on behalf of others; modern AI changes the machinery but not the need for human craft. That split—unchanged social skills alongside shifting production tools—is the practical problem for writers, editors and organisations.

A short, grounded history

The historical record is clear that ghostwriting is not a novel phenomenon. From early scribes to modern professional ghostwriters, people have written on behalf of others for centuries; recent overviews document how the role adapts as technologies change. Sources tracing that lineage show continuity in purpose even as methods evolve: the craft of shaping another person’s voice, ideas and authority is a social skill that rides successive waves of technology.

Why AI feels different

AI changes the machinery in three practical ways that make the experience of ghostwriting feel qualitatively different.

• Speed and scale: AI can generate drafts rapidly and at high volume, which shifts schedules and expectations for turnaround and iteration.
• Probabilistic production: Machine-generated text is the output of statistical processes, not a single, traceable human choice. That complicates traceability and makes provenance harder to read at a glance.
• Opacity of process: The internal mechanics of models and their training data are often opaque, so the social contract around authorship and responsibility becomes fuzzier.

Those differences alter how work feels, how risk is distributed and how audiences perceive authorship — they do not, however, remove the reasons people value a skilled ghostwriter.

What hasn’t gone away: the enduring craft

The qualities that make a ghostwriter valuable remain human and relational. Deep listening, editorial judgement and the ability to elicit, structure and shape lived experience are social capabilities that cannot be fully automated. Interviews and histories of the profession emphasise that successful ghostwriters have always combined subject-matter empathy with craft—skills that still anchor trust when AI is in the loop.

Practical guidance: treat AI as an instrument, redesign the sociotechnical system

For practitioners—freelancers, in-house writers, editors and product teams—the pragmatic path is to treat AI as a powerful instrument within redesigned agreements, workflows and interfaces. Below are concrete steps to do that.

• Redesign agreements:
  • Make AI use explicit. Include clauses that specify whether AI was used, and if so, which parts of the process it supported (brainstorming, first draft, editing, research assistance).
  • Clarify rights and attribution. Decide who owns outputs when models are used, and whether the named author takes full responsibility for content produced or merely curates it.
  • Define liability and revision windows. Agree who is responsible for factual accuracy, legal risk and corrections originating from model errors.
• Rework workflows:
  • Adopt human-in-the-loop checkpoints. Use AI for ideation and draft generation but require human review steps for tone, accuracy and voice before anything is final.
  • Keep iteration logs. Record prompts, model parameters and key edits so decisions are traceable and can be reviewed later.
  • Segment tasks by affordance. Use humans for elicitation, interviewing and narrative shaping; use AI for structured drafting, variant generation and routine rewrites.
• Design better interfaces:
  • Surface provenance metadata. Build or use tools that attach short provenance tags to drafts (e.g. “initial draft: model X, prompts A–C; revised by: editor Y”).
  • Provide edit-diff visibility. Ensure teams can see how AI output was edited so craft and judgement remain visible in the artefact.
  • Optimise prompts as part of the craft. Treat prompt design as a first-class editorial skill, and document effective prompts as part of team knowledge.
• Invest in people skills:
  • Train for elicitation and judgement. Workshops on interviewing, voice capture and editorial decision‑making preserve the human edge.
  • Build multidisciplinary teams. Combine writers, subject experts, legal advisers and UX folks to balance craft, accuracy and audience needs.

Operational notes for immediate adoption

Start small. Pilot a single content stream with explicit AI-use rules, logs and an author-review step. Capture lessons in a short playbook so the team can iterate governance alongside tooling. These low-friction experiments reveal the sociotechnical tensions—speed versus traceability, automation versus craft—so you can decide which trade-offs you are willing to accept.

Expanded from

This post expands on a point from Sociotechnical systems in software and AI: Ghostwriting has always existed — why does AI ghostwriting feel different?

Sources

• Original article — Sociotechnical systems in software and AI
• A history of ghostwriting — Charlotte Peacock
• A Brief History of Ghostwriting — E. Danielle Butler
• Interview With a Ghost (Writer) — The Scholarly Kitchen