While marketing from the AI companies vying for market share would have users believe that text-producing machines arrived ex nihilo, natural language generation represents an updated form of an ancient practice: ghostwriting. Debates about the ethics of this practice span millennia, but I want to focus on an argument between communications professors Ernest Bormann and Donald Smith from the 1961 issue of The Quarterly Journal of Speech about the ethics of ghostwriting. In what feels like a prescient anticipation of LLMs, they focus on the ability to believe the listed author was the originator of the idea as well as the labor of putting ideas into words articulately.
Bormann’s article draws from the English graduate students writing papers for athletes (262), presidential speeches ghostwriters (262), fraternity members sharing successful papers (263), and an undercover reporter who was paid to write college papers and even whole dissertations (265). In the course of these comparisons, Bormann directly calls out Smith’s position that “a college president may use a ghostwriter but a professor should write his own speeches” (265) as a “double standard.” Bormann distills this into a simple dilemma: if the speeches are too minor to be tended to by people in important positions, then the people in important positions should not waste their time giving those speeches; but if the speeches are too important to trust that the people in important positions will say the correct things appropriate to the situation, then the people in important positions probably should not hold those positions.
Smith responds to this article in the letters section, which also carries Bormann’s reply to Smith. The crux of the argument depends on whether or not ghostwriting is inherently deceptive. He begins, oddly enough, with “instances of such writing which are patently unethical, as in the case of the college student who turns in as his own work a paper written by a fraternity brother” (416-417) but does not offer evidence of the claim for being “patently unethical.” Rather, he insists that Bormann’s progression from this unethical act to other cases of unoriginal labor as unethical is jumping to conclusions, and he uses as his primary evidence a patriotic disbelief that a founding father of America be “in the same net with a plagiarist” (417). Smith moves on to address the existence of the “speech writer” career as a way in which the ghostwriting of presidential and other political speeches differs from the secrecy in which students submit other writers’ words as their own effort.
To shore up this argument, Smith asserts that the average American must certainly know that the president uses speech writers, unless they have “an aggressive level of ignorance” (418). For Smith, the office of the president is more important than the acknowledgement of the labor of ghostwriters, and assumptions about the knowledge of speechwriters suffice to compensate for their use for those in positions of authority, but leaves open the chance to investigate further: “We would be wise, I believe, in attacking the problem of ghostwriting … to seek to examine the variety of contexts within which such practice occurs, to appraise existing practice against the full range of purpose and necessity which it reflects, and to pinpoint our ethical judgements” (420). Bormann replies briefly in this letter section that Smith’s letter mischaracterizes much of what he says by ignoring distinctions in the original article and elevating the president and historical figures to a height that does not match the realities of their positions.
At stake in this conversation between Bormann and Smith, then, is the same issue many writing studies scholars are grappling with currently about the potentially detrimental use of LLMs in writing courses. Corporations seek to exploit AI to drive productivity, specifically pointing to the potential time-saving activity AI can provide (Brower 2025). This matches closely with what I found when researching the working lives of American corporate ghostwriters in marketing, technical writing, and public relations (Grimm 2022): the labor of writing is hidden behind various titles and completed with limited access to the listed author and more interaction with the marketing strategy materials. In this ghostwriting economy, text generation is what matters, not the authenticity of the listed author’s labor.
In 2024, Gallagher connected ghostwriting to AI use through his findings based on students’ ghostwriting experiences: students are asked to write “anonymously on behalf of the organization” (200) and “writing for or as someone else” (203) in professional capacities, much of which is being redirected to AI prompts (208-09). Text generation that was previously labeled as marketing, public relations, research assistants, etc., is now being completed by large-language model text generation without the need to hire peripheral employees and with minor human oversight.
While Smith claimed the widespread knowledge of speech writers was sufficient to cover any accusations of ghostwriting as an inherently deceptive practice, the common adaptation of AI into everyday life threatens to amplify this argument at ever-increasing volume. Because we all know it can be used, its use should be excused for those who cannot be bothered to write. Following Smith’s line of reasoning, signing a person’s name to a piece of AI-generated text is not deception anywhere except perhaps in education, where students and instructors are expected to represent their writing more earnestly.
Thus, it becomes the ghost-in-the-loop who may need to occupy our attention. According to Smith and AI proprietors, humans are too varied to expect a level playing field in any loops in which we find them: CEOs and politicians and college officials should be praised for passing off other people’s words as their own to protect their busy schedules; professors should own their knowledge creation without sacrificing their labor to the gods of efficiency; and students should do honest work without paying others for words or cranking out chunks of text when the process of doing the labor matters most to their thinking and learning.
As we craft our AI policies, we need to consider students’ learning first. When AI might hinder their learning, we should step in to reduce its likelihood to offload important tasks. When AI might provide a way to help learners process information, we should consider allowing students to interact with AI in transparent ways. This matches how we have always viewed writing to learn, and we should not let the newest version of ghostwriting cloud our view or forget that we’ve been having this conversation for decades, if not centuries or millennia. As long as we allow ourselves to stay in a haunted attic of humanity, let us focus on the ghosts in the loop.
References
Bormann, E. (1961). Ethics of ghostwritten speeches. Quarterly journal of speech, 47(4), pp. 262-268.
Bormann, E. (1961). Ghostwritten speeches – A reply. Letter. Quarterly journal of speech, 47(4), pp. 262-268.
Brower, T. (2025). “How AI is changing work and the human experience.” Forbes, 22 Dec. 2025. https://www.forbes.com/sites/tracybrower/2025/12/22/how-ai-is-changing-work-and-the-human-experience/
Gallagher, C. (2024). “‘This weird thing I’m discovering’: Toward a critical pedagogical approach to ghostwriting.” Pedagogy, 24(2), pp. 195-214.
Grimm, C. (2022). Ghostwriting as a critical lens: Authorship and attribution in professional and academic contexts. [Doctoral dissertation, Georgia State University]. Scholarworks. https://scholarworks.gsu.edu/items/b19f73a0-45ce-48a1-a65c-68eec66f3632
Smith, D. (1961). “Ghostwritten speeches.” Letter. Quarterly journal of speech, 47(4), pp. 416-420.
Generative AI Disclosure:
I did not use A.I. for the writing or research of this blog post. All errors and accuracies originated with me and are the fruit of my independent labor of thinking and writing.
