Keep Your Sense of Authorship When Writing With AI

Imagine asking AI to draft a story from an idea you care about. The result is fluent and recognisable: your setting, your characters, even something close to your preferred tone. Yet reading it produces an awkward question. How much of the finished piece feels shaped by you?

That reaction is not inevitable, and it does not mean AI assistance has failed. It points to a separate design question: how can a writing process give people useful help while preserving the involvement they value? Recent research suggests that opportunities to steer and revise matter to the experience of authorship.

What psychological ownership means here

Psychological ownership describes the subjective feeling that something is yours. In writing, that feeling can involve recognising your intentions and decisions in the finished text. It is a different question from whether the piece is accurate, effective or legally protected.

Keeping those questions separate is useful. A paragraph can sound polished while leaving its user feeling detached. A deeply personal draft can still need substantial improvement. Feeling involved is one outcome worth considering alongside the quality and purpose of the work.

What the recent experiment found

A preregistered experiment published online on 22 August 2026 by Niki Pennanen, Pyry Kanerva and Christian Guckelsberger involved 273 online participants. It compared unassisted short-story writing with AI-assisted conditions varying prompting style and whether editing was available.

Editing increased psychological ownership, especially following single-prompt generation. Iterative prompting increased ownership when editing was unavailable, but not when editing was available. Manual writing produced higher ownership than all the AI-assisted conditions.

These results concern particular short-story tasks. They cannot establish lasting effects on creative identity or what any individual will feel. They identify opportunities to influence text as worth investigating.

Feeling ownership and claiming authorship can diverge

Earlier research by Fiona Draxler and colleagues, reported in a 2024 journal article on the AI ghostwriter effect, examined two studies with 30 and 96 participants. Participants could feel limited ownership of generated writing while still presenting themselves as authors. Greater influence over the text increased their sense of ownership; personalising the generated text did not remove the reported effect.

The distinction is useful without turning it into an accusation. A byline, a familiar style and a person’s experience of contributing are separate things to examine. A text sounding like you does not answer every question about the process through which you arrived at it.

Decide which choices you want to keep

The following suggestions are an editorial workflow to try, rather than an intervention tested in those studies. Start by deciding which parts of the assignment you want to shape directly. For a meaningful story, that might be the conflict, the ending or the emotional distance between narrator and reader.

Write those intentions in a short note before asking for a complete draft. You do not need to manufacture a long brief. Three decisions can be enough to give your review a direction:

• Purpose: what you want the reader to understand, question or feel.
• Commitment: an idea, example or creative choice you want the piece to preserve.
• Open question: something you genuinely want help exploring.

Suppose a fictional writer wants a story about reconciliation that ends with uncertainty. They could ask for alternative ways to reveal the conflict, assess the suggestions and keep the unresolved ending. Their contribution is visible in the choice they defend, rather than in the length of the instructions.

Revise towards an intention

When a generated draft arrives, ask what you want to retain, reject or transform. Review the argument or story before concentrating on surface phrasing. Does the opening create the right expectation? Does the example support the intended point? Does the ending settle a question you wanted to leave open?

A useful revision has a reason. Replacing an abstract opening with a concrete scene might make the reader encounter the problem sooner. Removing a confident conclusion might preserve uncertainty that the evidence requires. Those are substantive decisions even when they affect relatively few words.

Avoid treating editing as a quota. The research does not supply a percentage of changed words that guarantees ownership. Nor should unnecessary rewriting become the price of using assistance. Keep changes that serve the work and the involvement you want.

Keep a small record of consequential choices

For an important piece, save the initial intention, a meaningful draft and a short note explaining a few final decisions. This gives you material to revisit when asking why the work took its final form.

MaryChuks’s authorship ledger guide offers a fuller approach to documenting creative inputs, selection and transformation. For a writing exercise, a few clear notes may be sufficient. The record should reflect what happened, including substantial contributions from the tool.

That record also helps when explaining your process to an editor or collaborator. Describe the assistance and your own decisions accurately, rather than relying on a vague claim that everything was either entirely manual or entirely generated.

Choose involvement to suit the task

There is no need to feel the same attachment to every piece of writing. Someone may welcome extensive drafting help for routine administration while wanting close involvement in fiction, a personal essay or a statement carrying their professional judgement.

For your next meaningful writing task, try keeping one important decision in your own hands and making one purposeful revision. Afterwards, ask two separate questions: did the piece achieve its purpose, and did the process give you the involvement you wanted?

Your answer may favour a different workflow next time. AI assistance can be arranged around the kind of participation you value. A useful writing partner leaves room for you to recognise, explain and stand behind the decisions that shaped the finished work.

Sources

• Pennanen, Kanerva and Guckelsberger, How prompting and editing shape psychological ownership in AI-assisted creative writing, Computers in Human Behavior: Artificial Humans, online publication 22 August 2026.
• Draxler and colleagues, The AI Ghostwriter Effect, ACM Transactions on Computer-Human Interaction, 2024 journal version; first published online in 2023.