I recently moderated a panel titled “Beyond Track Changes: Measuring the Impact of Human Manuscript Editing in an AI-Driven Publishing Landscape”, hosted by the Manuscript Editors Professional Role Group during the EASE (European Association of Science Editors) summer symposium. What I found particularly intriguing was the discussion on author voice and AI’s influence on it.

For over 24 years now, editors at Editage have received author instructions asking them to make their writing “more native” before resubmission. Such requests are common and have mostly been interpreted as a request for better English. But is that all? When we read a manuscript, we naturally infer qualities about the author based on the “voice” we encounter in the manuscript. The writing style, the confidence in the “author’s voice,” cultural context, the author’s familiarity (or lack thereof) with disciplinary norms, and their understanding of the fundamentals of manuscript preparation and submission are valuable signals. On the other hand, these same signals may introduce bias. A well-written paper may create a stronger first impression than a paper with science that is equally rigorous but with language that is less refined. Likewise, writing style may reveal an author’s career stage, geographic region, or educational background, even though none of these are reliable indicators of scientific quality.

Moreover, with AI coming into the picture, the lines of “native” versus “non-native writing” are getting blurred. What happens then to “author voice,” which has always been an important consideration to authors and reviewers alike? If AI is helping researchers with their writing, what does “author voice” really mean anymore?

In creative writing, the meaning of author voice is quite clear; it represents the author’s distinctive personal style of writing. But in academic writing does a personal style of writing even exist, with academic writing conventions being so standardized and authors being expected to follow journal guidelines and use correct terminology? Do we then refer to author voice as the choice of words used, or is it about the way the arguments are structured? Could it be the way the research question is framed? Or is it the rhythm or flow of the paper? Or is it just an intellectual style? Could author voice in academia then be defined as a reflection of disciplinary and cultural identity? Before we can ask whether AI threatens author voice, we first need to ask a more fundamental question: what exactly do we mean by “author voice” in scholarly writing?

The more I thought about this, the more elusive the idea became. Did a single identifiable definition of author voice ever truly exist in academia? There are so many layers to how we consider it. Think about multi-authorship. Each author brings in a different writing style, a different perspective, and a different idea and way of thinking. Is the “voice” then that of the first author who weaves these different perspectives together cohesively, or does it become something new, for instance a sort of “team voice”? I know that when I write, my voice feels distinctly my own versus when I edit someone else’s writing or contribute to a collaborative project. I write very differently than with the objective of building on someone else’s ideas. In such cases does author voice still belong to the author? Or is it something that emerges through collaboration?

Even for single author papers, does the voice really belong to the individual? Supervisors and mentors often help re-shape arguments. Peer reviewers may also suggest changes that can fundamentally change the structure of a paper. By the time the paper is finished and ready for publication, the first draft may bear little resemblance to the final one.  Since the author voice is getting influenced during the writing and submission process, does the voice of the paper itself evolve from the “author voice” to something broader, like a “published voice” of the paper that is shaped by authors, reviewers, mentors, and the conventions of the discipline itself?

If we look at it from this perspective, then AI is becoming another participant in the published voice of a paper. Will this help democratize research by eliminating barriers for those whose expertise lies in science and not the English language? Or more importantly will well-written papers begin to get mistaken for AI-generated ones? Is this a new form of bias that will slowly evolve?

The question of author voice therefore is not new. But with AI in the picture, this question is now impossible to ignore. Perhaps, AI has not changed author voice as much as it has exposed how little we have understood it all along.