Does user control over AI text shape feelings of ownership?

Draxler et al. find that the sense of ownership follows the user's influence on the text, not the personalization of the model. The abstract reads: "Personalization of AI-generated texts did not impact the AI Ghostwriter Effect, and higher levels of participants' influence on texts increased their sense of ownership." The introduction makes the same point, saying "Subjective control over the interaction and the content increases the sense of ownership." Two results sit side by side: one variable that moved ownership, and one that did not. The first is reported from Study 1 in the introduction.

The reasoning draws on prior work. Lehmann et al. found that the sense of authorship "positively correlates with the degree of influence over the AI contribution," and the authors start from that link. They separate objective control, the influence a user exerts through interaction methods named Writing, Editing, Choosing and Getting, from perceived control, the sense of being able to shape the text beforehand, and from leadership, the user's perceived initiative. H2.1 and H2.2 predict that influence affects control and leadership. For personalization, H3 predicts that ownership is independent of its quality. The authors compare fine-tuning with "placebo-personalization," where the AI is only labeled as personalized. Their basis is prior work finding that usability does not depend on personalization quality, and that non-adaptive systems are judged adaptive when introduced as such. The null result in the abstract is the pattern H3 predicted.

The closest library note, Does ownership framing change how much writers rely on AI?, treats ownership as a cause: owners lean on AI suggestions. This excerpt addresses what comes before ownership, namely how much the user shapes the text, and it rules out the model's personalization as the lever. The effect itself is described in Do people feel they own AI-generated text they use?. Read together, the two notes suggest a chain from influence to ownership to reliance. That chain is our reading; neither excerpt tests the full path.

The excerpt does not say which interaction method produced the most ownership, how influence was scored, or what the effect sizes were. It reports no interface test, though the authors say understanding control, ownership and authorship "informs the interaction design of future AI-supported text-generation systems." At the strength the evidence allows, giving users more say over generated text looks like a plausible lever on ownership, and a personalization claim alone may not supply one. That is a hypothesis for interface testing, and the excerpt does not establish it.